Compare commits
No commits in common. "main" and "0.13.0" have entirely different histories.
@ -0,0 +1,25 @@
|
||||
package com.knecon.fforesight.service.ocr.v1.api.model;
|
||||
|
||||
import java.util.ArrayList;
|
||||
import java.util.List;
|
||||
|
||||
import lombok.AccessLevel;
|
||||
import lombok.AllArgsConstructor;
|
||||
import lombok.Builder;
|
||||
import lombok.Getter;
|
||||
import lombok.experimental.FieldDefaults;
|
||||
|
||||
@Getter
|
||||
@Builder
|
||||
@AllArgsConstructor
|
||||
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
|
||||
public class AzureAnalyzeResult {
|
||||
|
||||
@Builder.Default
|
||||
List<KeyValuePair> keyValuePairs = new ArrayList<>();
|
||||
@Builder.Default
|
||||
List<TextRegion> handWrittenText = new ArrayList<>();
|
||||
@Builder.Default
|
||||
List<Figure> figures = new ArrayList<>();
|
||||
|
||||
}
|
||||
@ -73,10 +73,4 @@ public class DocumentRequest {
|
||||
return Optional.ofNullable(originDocumentId);
|
||||
}
|
||||
|
||||
|
||||
public Set<AzureOcrFeature> getFeatures() {
|
||||
|
||||
return features == null ? Collections.emptySet() : features;
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
@ -1,11 +1,10 @@
|
||||
package com.knecon.fforesight.service.ocr.v1.api.model;
|
||||
|
||||
import java.util.List;
|
||||
import java.util.Optional;
|
||||
|
||||
import lombok.Builder;
|
||||
|
||||
@Builder
|
||||
public record Figure(TextRegion caption, Region image, List<TextRegion> footnotes) {
|
||||
public record Figure(Optional<TextRegion> caption, Region image) {
|
||||
|
||||
}
|
||||
|
||||
@ -1,23 +0,0 @@
|
||||
package com.knecon.fforesight.service.ocr.v1.api.model;
|
||||
|
||||
import java.util.Collections;
|
||||
import java.util.LinkedList;
|
||||
import java.util.List;
|
||||
|
||||
public record IdpResult(List<KeyValuePair> keyValuePairs, List<TextRegion> handWrittenText, List<Figure> figures, List<Table> tables) {
|
||||
|
||||
public static IdpResult initSynchronized() {
|
||||
|
||||
return new IdpResult(Collections.synchronizedList(new LinkedList<>()),
|
||||
Collections.synchronizedList(new LinkedList<>()),
|
||||
Collections.synchronizedList(new LinkedList<>()),
|
||||
Collections.synchronizedList(new LinkedList<>()));
|
||||
}
|
||||
|
||||
|
||||
public static IdpResult empty() {
|
||||
|
||||
return new IdpResult(Collections.emptyList(), Collections.emptyList(), Collections.emptyList(), Collections.emptyList());
|
||||
}
|
||||
|
||||
}
|
||||
@ -1,8 +1,5 @@
|
||||
package com.knecon.fforesight.service.ocr.v1.api.model;
|
||||
|
||||
import java.util.Collections;
|
||||
import java.util.Set;
|
||||
|
||||
import lombok.AllArgsConstructor;
|
||||
import lombok.Builder;
|
||||
import lombok.Data;
|
||||
@ -15,16 +12,9 @@ import lombok.NoArgsConstructor;
|
||||
public class OCRStatusUpdateResponse {
|
||||
|
||||
private String fileId;
|
||||
private Set<AzureOcrFeature> features;
|
||||
private int numberOfPagesToOCR;
|
||||
private int numberOfOCRedPages;
|
||||
private boolean ocrFinished;
|
||||
private boolean ocrStarted;
|
||||
|
||||
|
||||
public Set<AzureOcrFeature> getFeatures() {
|
||||
|
||||
return features == null ? Collections.emptySet() : features;
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
@ -5,12 +5,9 @@ import java.awt.geom.Line2D;
|
||||
import java.awt.geom.Point2D;
|
||||
import java.awt.geom.Rectangle2D;
|
||||
import java.util.List;
|
||||
import java.util.Objects;
|
||||
import java.util.stream.Stream;
|
||||
|
||||
import lombok.Getter;
|
||||
|
||||
public final class QuadPoint {
|
||||
public record QuadPoint(Point2D a, Point2D b, Point2D c, Point2D d) {
|
||||
|
||||
public enum Direction {
|
||||
RIGHT,
|
||||
@ -44,38 +41,7 @@ public final class QuadPoint {
|
||||
* ?|_____|?
|
||||
*/
|
||||
}
|
||||
|
||||
private static final double THRESHOLD_ANGLE = Math.toRadians(5); // QuadPoint is considered straight, when its angles are below this threshold.
|
||||
|
||||
private final Point2D a;
|
||||
private final Point2D b;
|
||||
private final Point2D c;
|
||||
private final Point2D d;
|
||||
@Getter
|
||||
private final Direction direction;
|
||||
|
||||
|
||||
// This constructor assumes, the points form a convex polygon, I will omit the assertion for performance reasons.
|
||||
public QuadPoint(Point2D a, Point2D b, Point2D c, Point2D d) {
|
||||
|
||||
this.a = a;
|
||||
this.b = b;
|
||||
this.c = c;
|
||||
this.d = d;
|
||||
this.direction = calculateDirection();
|
||||
}
|
||||
|
||||
|
||||
private Direction calculateDirection() {
|
||||
|
||||
if (isHorizontal()) {
|
||||
return a.getX() < d.getX() ? Direction.RIGHT : Direction.LEFT;
|
||||
}
|
||||
if (isVertical()) {
|
||||
return a.getY() < d.getY() ? Direction.UP : Direction.DOWN;
|
||||
}
|
||||
return Direction.NONE;
|
||||
}
|
||||
private static final double THRESHOLD_ANGLE = Math.toRadians(5); // QuadPoint is considered straight, when its angles are below this threshold
|
||||
|
||||
|
||||
public static QuadPoint fromRectangle2D(Rectangle2D rectangle2D) {
|
||||
@ -103,9 +69,7 @@ public final class QuadPoint {
|
||||
|
||||
public static QuadPoint fromPolygons(List<Double> polygon) {
|
||||
|
||||
if (polygon.size() != 8) {
|
||||
throw new AssertionError();
|
||||
}
|
||||
assert polygon.size() == 8;
|
||||
return new QuadPoint(new Point2D.Double(polygon.get(0), polygon.get(1)),
|
||||
new Point2D.Double(polygon.get(6), polygon.get(7)),
|
||||
new Point2D.Double(polygon.get(4), polygon.get(5)),
|
||||
@ -151,6 +115,18 @@ public final class QuadPoint {
|
||||
}
|
||||
|
||||
|
||||
public Direction getDirection() {
|
||||
|
||||
if (isHorizontal()) {
|
||||
return a.getX() < d.getX() ? Direction.RIGHT : Direction.LEFT;
|
||||
}
|
||||
if (isVertical()) {
|
||||
return a.getY() < d.getY() ? Direction.UP : Direction.DOWN;
|
||||
}
|
||||
return Direction.NONE;
|
||||
}
|
||||
|
||||
|
||||
public Stream<Line2D> asLines() {
|
||||
|
||||
return Stream.of(new Line2D.Double(a(), b()), new Line2D.Double(b(), c()), new Line2D.Double(c(), d()), new Line2D.Double(d(), a()));
|
||||
@ -170,142 +146,6 @@ public final class QuadPoint {
|
||||
}
|
||||
|
||||
|
||||
public boolean contains(double x, double y) {
|
||||
// split into two triangles, test if either contains the point, assumes the QuadPoint is convex and created correctly. More specifically, the points must be in the correct order.
|
||||
return triangleContains(a, b, c, x, y) || triangleContains(a, c, d, x, y);
|
||||
}
|
||||
|
||||
|
||||
/*
|
||||
checks if a triangle contains a point by converting the point to barycentric coordinates using cramer's rule and then checking if the linear combination is within the bounds of the triangle.
|
||||
https://en.wikipedia.org/wiki/Barycentric_coordinate_system#Barycentric_coordinates_on_triangles
|
||||
*/
|
||||
private boolean triangleContains(Point2D a, Point2D b, Point2D c, double x, double y) {
|
||||
|
||||
// area of the triangle
|
||||
double denominator = ((b.getY() - c.getY()) * (a.getX() - c.getX()) + (c.getX() - b.getX()) * (a.getY() - c.getY()));
|
||||
double invertedDenominator = 1.0 / denominator;
|
||||
double alpha = ((b.getY() - c.getY()) * (x - c.getX()) + (c.getX() - b.getX()) * (y - c.getY())) * invertedDenominator;
|
||||
double beta = ((c.getY() - a.getY()) * (x - c.getX()) + (a.getX() - c.getX()) * (y - c.getY())) * invertedDenominator;
|
||||
|
||||
return alpha >= 0 && beta >= 0 && alpha + beta <= 1;
|
||||
}
|
||||
|
||||
|
||||
public boolean contains(Point2D p) {
|
||||
|
||||
return contains(p.getX(), p.getY());
|
||||
}
|
||||
|
||||
|
||||
public boolean contains(Rectangle2D r) {
|
||||
|
||||
double x = r.getX();
|
||||
double y = r.getY();
|
||||
double maxY = r.getMaxY();
|
||||
double maxX = r.getMaxX();
|
||||
|
||||
Point2D p1 = new Point2D.Double(x, y);
|
||||
Point2D p2 = new Point2D.Double(x, maxY);
|
||||
Point2D p3 = new Point2D.Double(maxX, maxY);
|
||||
Point2D p4 = new Point2D.Double(maxX, y);
|
||||
|
||||
return contains(p1) && contains(p2) && contains(p3) && contains(p4);
|
||||
}
|
||||
|
||||
|
||||
public double getCenterX() {
|
||||
|
||||
return (a.getX() + b.getX() + c.getX() + d.getX()) / 4;
|
||||
}
|
||||
|
||||
|
||||
public double getCenterY() {
|
||||
|
||||
return (a.getY() + b.getY() + c.getY() + d.getY()) / 4;
|
||||
}
|
||||
|
||||
|
||||
public Point2D getCenter() {
|
||||
|
||||
return new Point2D.Double(getCenterX(), getCenterY());
|
||||
}
|
||||
|
||||
|
||||
public boolean intersects(Line2D line) {
|
||||
|
||||
return contains(line.getP1()) || contains(line.getP2()) || asLines().anyMatch(qLine -> qLine.intersectsLine(line));
|
||||
}
|
||||
|
||||
|
||||
public Line2D getRightLine() {
|
||||
|
||||
return new Line2D.Double(getTopRight(), getLowerRight());
|
||||
}
|
||||
|
||||
|
||||
public Line2D getLeftLine() {
|
||||
|
||||
return new Line2D.Double(getTopLeft(), getLowerLeft());
|
||||
}
|
||||
|
||||
|
||||
public Line2D getBottomLine() {
|
||||
|
||||
return new Line2D.Double(getLowerLeft(), getLowerRight());
|
||||
}
|
||||
|
||||
|
||||
public Line2D getTopLine() {
|
||||
|
||||
return new Line2D.Double(getTopLeft(), getTopRight());
|
||||
}
|
||||
|
||||
|
||||
public Point2D getTopLeft() {
|
||||
|
||||
return switch (direction) {
|
||||
case DOWN -> a;
|
||||
case LEFT -> d;
|
||||
case UP -> c;
|
||||
default -> b;
|
||||
};
|
||||
}
|
||||
|
||||
|
||||
public Point2D getTopRight() {
|
||||
|
||||
return switch (direction) {
|
||||
case DOWN -> b;
|
||||
case LEFT -> a;
|
||||
case UP -> d;
|
||||
default -> c;
|
||||
};
|
||||
}
|
||||
|
||||
|
||||
public Point2D getLowerRight() {
|
||||
|
||||
return switch (direction) {
|
||||
case DOWN -> c;
|
||||
case LEFT -> b;
|
||||
case UP -> a;
|
||||
default -> d;
|
||||
};
|
||||
}
|
||||
|
||||
|
||||
public Point2D getLowerLeft() {
|
||||
|
||||
return switch (direction) {
|
||||
case DOWN -> d;
|
||||
case LEFT -> c;
|
||||
case UP -> b;
|
||||
default -> a;
|
||||
};
|
||||
}
|
||||
|
||||
|
||||
/**
|
||||
* Determines if the given QuadPoint aligns with this QuadPoint within a given threshold.
|
||||
* It does os by trying every possible combination of aligning sides. It starts with the most likely combination of ab and cd.
|
||||
@ -384,37 +224,4 @@ public final class QuadPoint {
|
||||
return Math.atan2(deltaY, deltaX);
|
||||
}
|
||||
|
||||
|
||||
public Point2D a() {return a;}
|
||||
|
||||
|
||||
public Point2D b() {return b;}
|
||||
|
||||
|
||||
public Point2D c() {return c;}
|
||||
|
||||
|
||||
public Point2D d() {return d;}
|
||||
|
||||
|
||||
@Override
|
||||
public boolean equals(Object obj) {
|
||||
|
||||
if (obj == this) {
|
||||
return true;
|
||||
}
|
||||
if (obj == null || obj.getClass() != this.getClass()) {
|
||||
return false;
|
||||
}
|
||||
var that = (QuadPoint) obj;
|
||||
return Objects.equals(this.a, that.a) && Objects.equals(this.b, that.b) && Objects.equals(this.c, that.c) && Objects.equals(this.d, that.d);
|
||||
}
|
||||
|
||||
|
||||
@Override
|
||||
public int hashCode() {
|
||||
|
||||
return Objects.hash(a, b, c, d);
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
@ -5,9 +5,4 @@ import lombok.Builder;
|
||||
@Builder
|
||||
public record QuadPointData(float[] values) {
|
||||
|
||||
public QuadPoint get() {
|
||||
|
||||
return QuadPoint.fromData(this);
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
@ -1,7 +0,0 @@
|
||||
package com.knecon.fforesight.service.ocr.v1.api.model;
|
||||
|
||||
import java.util.List;
|
||||
|
||||
public record Table(TextRegion caption, int numberOfCols, int numberOfRows, List<TableCell> cells, List<TextRegion> footnotes, List<Region> bboxes) {
|
||||
|
||||
}
|
||||
@ -1,5 +0,0 @@
|
||||
package com.knecon.fforesight.service.ocr.v1.api.model;
|
||||
|
||||
public record TableCell(TextRegion textRegion, int row, int col, TableCellType kind) {
|
||||
|
||||
}
|
||||
@ -1,5 +0,0 @@
|
||||
package com.knecon.fforesight.service.ocr.v1.api.model;
|
||||
|
||||
public enum TableCellType {
|
||||
ROW_HEADER, COLUMN_HEADER, CONTENT, STUB_HEAD, DESCRIPTION
|
||||
}
|
||||
@ -10,18 +10,19 @@ configurations {
|
||||
}
|
||||
|
||||
dependencies {
|
||||
implementation(project(":azure-ocr-service-api"))
|
||||
implementation("net.sourceforge.tess4j:tess4j:5.8.0")
|
||||
implementation("com.iqser.red.commons:metric-commons:2.1.0")
|
||||
implementation("com.pdftron:PDFNet:11.0.0")
|
||||
implementation("org.apache.pdfbox:pdfbox:3.0.0")
|
||||
implementation("org.apache.commons:commons-math3:3.6.1")
|
||||
implementation("com.amazonaws:aws-java-sdk-kms:1.12.440")
|
||||
implementation("com.google.guava:guava:31.1-jre")
|
||||
implementation("com.knecon.fforesight:viewer-doc-processor:0.193.0")
|
||||
implementation("com.azure:azure-ai-documentintelligence:1.0.0")
|
||||
|
||||
implementation("com.iqser.red.commons:pdftron-logic-commons:2.32.0")
|
||||
|
||||
api(project(":azure-ocr-service-api"))
|
||||
api("com.iqser.red.service:persistence-service-internal-api-v1:2.224.0")
|
||||
api("net.sourceforge.tess4j:tess4j:5.8.0")
|
||||
api("com.iqser.red.commons:metric-commons:2.1.0")
|
||||
api("com.iqser.red.commons:storage-commons:2.49.0")
|
||||
api("com.knecon.fforesight:tenant-commons:0.30.0")
|
||||
api("com.pdftron:PDFNet:10.7.0")
|
||||
api("org.apache.pdfbox:pdfbox:3.0.0")
|
||||
api("org.apache.commons:commons-math3:3.6.1")
|
||||
api("com.amazonaws:aws-java-sdk-kms:1.12.440")
|
||||
api("com.google.guava:guava:31.1-jre")
|
||||
api("com.knecon.fforesight:viewer-doc-processor:0.177.0")
|
||||
api("com.azure:azure-ai-documentintelligence:1.0.0-beta.3")
|
||||
api("com.iqser.red.commons:pdftron-logic-commons:2.31.0")
|
||||
testImplementation("org.junit.jupiter:junit-jupiter:5.8.1")
|
||||
}
|
||||
|
||||
@ -6,8 +6,6 @@ import org.springframework.context.annotation.Bean;
|
||||
import org.springframework.context.annotation.ComponentScan;
|
||||
import org.springframework.context.annotation.Configuration;
|
||||
|
||||
import com.iqser.red.pdftronlogic.commons.InvisibleElementRemovalService;
|
||||
import com.iqser.red.pdftronlogic.commons.WatermarkRemovalService;
|
||||
import com.knecon.fforesight.service.viewerdoc.service.PDFTronViewerDocumentService;
|
||||
|
||||
import io.micrometer.observation.ObservationRegistry;
|
||||
@ -24,18 +22,4 @@ public class OcrServiceProcessorConfiguration {
|
||||
return new PDFTronViewerDocumentService(registry);
|
||||
}
|
||||
|
||||
|
||||
@Bean
|
||||
public InvisibleElementRemovalService invisibleElementRemovalService() {
|
||||
|
||||
return new InvisibleElementRemovalService();
|
||||
}
|
||||
|
||||
|
||||
@Bean
|
||||
public WatermarkRemovalService watermarkRemovalService() {
|
||||
|
||||
return new WatermarkRemovalService();
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
@ -11,16 +11,14 @@ import lombok.experimental.FieldDefaults;
|
||||
@FieldDefaults(level = AccessLevel.PRIVATE)
|
||||
public class OcrServiceSettings {
|
||||
|
||||
// Limits the number of concurrent calls to azure
|
||||
int concurrency = 2;
|
||||
// Limits the number of concurrent calls to the azure API. In my very rudimentary testing, azure starts throwing "too many requests" errors at around 80/s. Higher numbers greatly improve the speed.
|
||||
int concurrency = 8;
|
||||
// Limits the number of pages per call.
|
||||
int batchSize = 32;
|
||||
int batchSize = 128;
|
||||
|
||||
boolean debug; // writes the ocr layer visibly to the viewer doc pdf
|
||||
boolean drawTablesAsLines; // writes the tables to the PDF as invisible lines.
|
||||
boolean snuggify = true; // attempts to shrink the word boxes returned by azure to fit the actual word pixels snug
|
||||
boolean useCaches; // skips azure api, pdf rendering and image processing, when the files are already present
|
||||
boolean azureFontStyleDetection; // omits all image processing and uses azures FONT_STYLE feature (costs 0.6ct per page)
|
||||
boolean snuggify = true; // Enables bold detection using ghostscript and leptonica
|
||||
String contentFormat; // Either markdown or text. But, for whatever reason, with markdown enabled, key-values are not written by azure....
|
||||
|
||||
}
|
||||
|
||||
@ -21,8 +21,6 @@ public class NativeLibrariesInitializer {
|
||||
@Value("${pdftron.license:}")
|
||||
private String pdftronLicense;
|
||||
|
||||
@Value("${native-libs.path:}")
|
||||
private String nativeLibsPath;
|
||||
|
||||
@SneakyThrows
|
||||
@PostConstruct
|
||||
@ -34,8 +32,8 @@ public class NativeLibrariesInitializer {
|
||||
PDFNet.setTempPath("/tmp/pdftron");
|
||||
PDFNet.initialize(pdftronLicense);
|
||||
|
||||
log.info("Setting jna.library.path: {}", nativeLibsPath);
|
||||
System.setProperty("jna.library.path", nativeLibsPath);
|
||||
log.info("Setting jna.library.path: {}", System.getenv("VCPKG_DYNAMIC_LIB"));
|
||||
System.setProperty("jna.library.path", System.getenv("VCPKG_DYNAMIC_LIB"));
|
||||
|
||||
log.info("Asserting Native Libraries loaded");
|
||||
|
||||
|
||||
@ -1,102 +0,0 @@
|
||||
package com.knecon.fforesight.service.ocr.processor.model;
|
||||
|
||||
import java.util.ArrayList;
|
||||
import java.util.Collections;
|
||||
import java.util.Comparator;
|
||||
import java.util.List;
|
||||
import java.util.Map;
|
||||
import java.util.TreeMap;
|
||||
import java.util.stream.Collectors;
|
||||
|
||||
import com.azure.ai.documentintelligence.models.AnalyzeResult;
|
||||
import com.azure.ai.documentintelligence.models.DocumentPage;
|
||||
import com.azure.ai.documentintelligence.models.DocumentSpan;
|
||||
import com.azure.ai.documentintelligence.models.DocumentWord;
|
||||
|
||||
public class DocumentSpanLookup {
|
||||
|
||||
List<PageSpanLookup> documentWordLookup;
|
||||
|
||||
|
||||
public DocumentSpanLookup(AnalyzeResult analyzeResult) {
|
||||
|
||||
documentWordLookup = new ArrayList<>(analyzeResult.getPages().size());
|
||||
int offset = 0;
|
||||
for (DocumentPage page : analyzeResult.getPages()) {
|
||||
|
||||
if (page.getWords() == null || page.getWords().isEmpty()) {
|
||||
documentWordLookup.add(new PageSpanLookup(offset, offset, null));
|
||||
}
|
||||
int start = page.getWords()
|
||||
.get(0).getSpan().getOffset();
|
||||
DocumentSpan span = page.getWords()
|
||||
.get(page.getWords().size() - 1).getSpan();
|
||||
int end = span.getOffset() + span.getLength();
|
||||
SpanLookup<DocumentWord> pageWords = new SpanLookup<>(page.getWords()
|
||||
.stream(), DocumentWord::getSpan);
|
||||
documentWordLookup.add(new PageSpanLookup(start, end, pageWords));
|
||||
offset = end + 1;
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
public List<WordOnPage> findWordsOnPages(DocumentSpan documentSpan) {
|
||||
|
||||
if (documentSpan == null) {
|
||||
return Collections.emptyList();
|
||||
}
|
||||
int firstSmallerIdx = findIdxOfFirstSmallerObject(documentSpan);
|
||||
PageSpanLookup firstPage = documentWordLookup.get(firstSmallerIdx);
|
||||
List<WordOnPage> wordsOnPages = new ArrayList<>();
|
||||
for (int pageNumber = firstSmallerIdx; pageNumber < documentWordLookup.size(); pageNumber++) {
|
||||
PageSpanLookup page = documentWordLookup.get(pageNumber);
|
||||
if (page.end >= documentSpan.getOffset()) {
|
||||
break;
|
||||
}
|
||||
firstPage.wordSpanLookup.findElementsContainedInSpan(documentSpan)
|
||||
.stream()
|
||||
.map(documentWord -> new WordOnPage(documentWord, firstSmallerIdx))
|
||||
.forEach(wordsOnPages::add);
|
||||
}
|
||||
return wordsOnPages;
|
||||
}
|
||||
|
||||
|
||||
private int findIdxOfFirstSmallerObject(DocumentSpan documentSpan) {
|
||||
|
||||
int idx = Collections.binarySearch(documentWordLookup, new PageSpanLookup(documentSpan.getOffset(), -1, null), Comparator.comparing(PageSpanLookup::start));
|
||||
|
||||
if (idx >= 0) {
|
||||
return idx;
|
||||
} else {
|
||||
int insertionPoint = -(idx + 1);
|
||||
|
||||
if (insertionPoint == 0) {
|
||||
return -1;
|
||||
}
|
||||
var lastSmaller = documentWordLookup.get(insertionPoint - 1);
|
||||
for (int resultIdx = insertionPoint - 2; resultIdx >= 0; resultIdx--) {
|
||||
if (documentWordLookup.get(resultIdx).compareTo(lastSmaller) == 0) {
|
||||
return resultIdx + 1;
|
||||
}
|
||||
}
|
||||
return 0;
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
public record WordOnPage(DocumentWord documentWord, int pageNumber) {
|
||||
|
||||
}
|
||||
|
||||
private record PageSpanLookup(int start, int end, SpanLookup<DocumentWord> wordSpanLookup) implements Comparable<PageSpanLookup> {
|
||||
|
||||
@Override
|
||||
public int compareTo(PageSpanLookup o) {
|
||||
|
||||
return Integer.compare(start, o.start);
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
}
|
||||
@ -1,7 +1,5 @@
|
||||
package com.knecon.fforesight.service.ocr.processor.model;
|
||||
|
||||
import java.io.File;
|
||||
|
||||
import net.sourceforge.lept4j.Leptonica1;
|
||||
import net.sourceforge.lept4j.Pix;
|
||||
|
||||
@ -12,10 +10,4 @@ public record ImageFile(int pageNumber, String absoluteFilePath) {
|
||||
return Leptonica1.pixRead(absoluteFilePath);
|
||||
}
|
||||
|
||||
|
||||
public boolean exists() {
|
||||
|
||||
return new File(absoluteFilePath).exists();
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
@ -2,25 +2,16 @@ package com.knecon.fforesight.service.ocr.processor.model;
|
||||
|
||||
import static com.knecon.fforesight.service.ocr.processor.utils.ListSplittingUtils.formatIntervals;
|
||||
|
||||
import java.io.File;
|
||||
import java.io.FileInputStream;
|
||||
import java.nio.file.Path;
|
||||
import java.util.ArrayList;
|
||||
import java.util.List;
|
||||
import java.util.function.Consumer;
|
||||
|
||||
import com.azure.ai.documentintelligence.models.AnalyzeResult;
|
||||
import com.azure.core.util.BinaryData;
|
||||
import com.azure.json.JsonOptions;
|
||||
import com.azure.json.JsonReader;
|
||||
import com.azure.json.implementation.DefaultJsonReader;
|
||||
import com.knecon.fforesight.service.ocr.processor.service.imageprocessing.GhostScriptService;
|
||||
|
||||
import lombok.AccessLevel;
|
||||
import lombok.Getter;
|
||||
import lombok.NonNull;
|
||||
import lombok.RequiredArgsConstructor;
|
||||
import lombok.SneakyThrows;
|
||||
import lombok.experimental.FieldDefaults;
|
||||
|
||||
@RequiredArgsConstructor
|
||||
@ -30,101 +21,23 @@ public final class PageBatch implements Comparable<PageBatch> {
|
||||
@Getter
|
||||
int index;
|
||||
@NonNull
|
||||
List<Integer> batchPageToOriginPageLookup;
|
||||
List<Integer> lookup;
|
||||
@NonNull
|
||||
@Getter
|
||||
Path batchDoc;
|
||||
@NonNull
|
||||
@Getter
|
||||
Path batchDir;
|
||||
|
||||
|
||||
@SneakyThrows
|
||||
public AnalyzeResult getAzureResultCache() {
|
||||
|
||||
try (var in = new FileInputStream(getAzureResultCacheFile()); JsonReader reader = DefaultJsonReader.fromStream(in, new JsonOptions());) {
|
||||
return AnalyzeResult.fromJson(reader);
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@SneakyThrows
|
||||
public File getAzureResultCacheFile() {
|
||||
|
||||
return batchDir.resolve("analyzeResult.json").toFile();
|
||||
}
|
||||
|
||||
|
||||
public List<ImageFile> getRenderedImageFiles() {
|
||||
|
||||
List<ImageFile> renderedImageFiles = new ArrayList<>();
|
||||
for (int i = 0; i < batchPageToOriginPageLookup.size(); i++) {
|
||||
renderedImageFiles.add(getRenderedImageFile(batchPageToOriginPageLookup.get(i), i + 1));
|
||||
}
|
||||
return renderedImageFiles;
|
||||
}
|
||||
|
||||
|
||||
public ImageFile getRenderedImageFile(int pageNumber, int numberInBatch) {
|
||||
|
||||
return new ImageFile(pageNumber, getRenderedImageNameFormat().formatted(numberInBatch));
|
||||
}
|
||||
|
||||
|
||||
public ImageFile getProcessedImageFile(int pageNumber, int numberInBatch) {
|
||||
|
||||
return new ImageFile(pageNumber, getProcessedImageNameFormat().formatted(numberInBatch));
|
||||
}
|
||||
|
||||
|
||||
public List<ImageFile> getProcessedImageFiles() {
|
||||
|
||||
List<ImageFile> processedImageFiles = new ArrayList<>();
|
||||
for (int i = 0; i < batchPageToOriginPageLookup.size(); i++) {
|
||||
processedImageFiles.add(getProcessedImageFile(batchPageToOriginPageLookup.get(i), i + 1));
|
||||
}
|
||||
return processedImageFiles;
|
||||
}
|
||||
|
||||
|
||||
public String getRenderedImageNameFormat() {
|
||||
|
||||
return getRenderedImageDir().resolve(getImageFormat()).toFile().toString();
|
||||
}
|
||||
|
||||
|
||||
public String getProcessedImageNameFormat() {
|
||||
|
||||
return getProcessedImageDir().resolve(getImageFormat()).toFile().toString();
|
||||
}
|
||||
|
||||
|
||||
private String getImageFormat() {
|
||||
|
||||
return "output_" + index + ".%04d" + GhostScriptService.FORMAT;
|
||||
}
|
||||
|
||||
|
||||
public Path getRenderedImageDir() {
|
||||
|
||||
return batchDir.resolve("rendered");
|
||||
}
|
||||
|
||||
|
||||
public Path getProcessedImageDir() {
|
||||
|
||||
return batchDir.resolve("processed");
|
||||
}
|
||||
Path imagePipelineDir;
|
||||
|
||||
|
||||
@Override
|
||||
public String toString() {
|
||||
|
||||
if (size() == 1) {
|
||||
return String.format("%d", batchPageToOriginPageLookup.get(0));
|
||||
return String.format("%d", lookup.get(0));
|
||||
}
|
||||
|
||||
List<String> intervals = formatIntervals(batchPageToOriginPageLookup);
|
||||
List<String> intervals = formatIntervals(lookup);
|
||||
if (intervals.size() > 4) {
|
||||
intervals = intervals.subList(0, 4);
|
||||
intervals.add("...");
|
||||
@ -136,46 +49,46 @@ public final class PageBatch implements Comparable<PageBatch> {
|
||||
|
||||
public void forEach(Consumer<? super Integer> consumer) {
|
||||
|
||||
batchPageToOriginPageLookup.forEach(consumer);
|
||||
lookup.forEach(consumer);
|
||||
}
|
||||
|
||||
|
||||
public List<Integer> getAllPageNumbers() {
|
||||
|
||||
return batchPageToOriginPageLookup;
|
||||
return lookup;
|
||||
}
|
||||
|
||||
|
||||
public int size() {
|
||||
|
||||
return batchPageToOriginPageLookup.size();
|
||||
return lookup.size();
|
||||
}
|
||||
|
||||
|
||||
public boolean isEmpty() {
|
||||
|
||||
return batchPageToOriginPageLookup.isEmpty();
|
||||
return lookup.isEmpty();
|
||||
}
|
||||
|
||||
|
||||
public int getPageNumber(int pageNumber) {
|
||||
|
||||
return batchPageToOriginPageLookup.get(pageNumber - 1);
|
||||
return lookup.get(pageNumber - 1);
|
||||
}
|
||||
|
||||
|
||||
@Override
|
||||
public int compareTo(PageBatch o) {
|
||||
|
||||
if (batchPageToOriginPageLookup.isEmpty() && o.batchPageToOriginPageLookup.isEmpty()) {
|
||||
if (lookup.isEmpty() && o.lookup.isEmpty()) {
|
||||
return 0;
|
||||
} else if (batchPageToOriginPageLookup.isEmpty()) {
|
||||
} else if (lookup.isEmpty()) {
|
||||
return 1;
|
||||
} else if (o.batchPageToOriginPageLookup.isEmpty()) {
|
||||
} else if (o.lookup.isEmpty()) {
|
||||
return -1;
|
||||
}
|
||||
|
||||
return Integer.compare(batchPageToOriginPageLookup.get(0), o.batchPageToOriginPageLookup.get(0));
|
||||
return Integer.compare(lookup.get(0), o.lookup.get(0));
|
||||
}
|
||||
|
||||
|
||||
|
||||
@ -13,7 +13,7 @@ import com.pdftron.pdf.Rect;
|
||||
|
||||
import lombok.SneakyThrows;
|
||||
|
||||
public record PageInformation(Rectangle2D mediabox, Rectangle2D cropBox, int number, int rotationDegrees, List<Rectangle2D> wordBBoxes) {
|
||||
public record PageInformation(Rectangle2D mediabox, int number, int rotationDegrees, List<Rectangle2D> wordBBoxes) {
|
||||
|
||||
@SneakyThrows
|
||||
public static Map<Integer, PageInformation> fromPDFDoc(PDFDoc pdfDoc) {
|
||||
@ -34,9 +34,8 @@ public record PageInformation(Rectangle2D mediabox, Rectangle2D cropBox, int num
|
||||
@SneakyThrows
|
||||
public static PageInformation fromPage(int pageNum, Page page) {
|
||||
|
||||
try (Rect mediaBox = page.getCropBox(); Rect cropBox = page.getCropBox()) {
|
||||
try (Rect mediaBox = page.getCropBox()) {
|
||||
return new PageInformation(new Rectangle2D.Double(mediaBox.getX1(), mediaBox.getY1(), mediaBox.getWidth(), mediaBox.getHeight()),
|
||||
new Rectangle2D.Double(cropBox.getX1(), cropBox.getY1(), cropBox.getWidth(), cropBox.getHeight()),
|
||||
pageNum,
|
||||
page.getRotation() * 90,
|
||||
DocumentTextExtractor.getTextBBoxes(page));
|
||||
|
||||
@ -162,7 +162,7 @@ public class Statistics {
|
||||
|
||||
return batchStats.values()
|
||||
.stream()
|
||||
.mapToLong(BatchStats::getMappingResultDuration)
|
||||
.mapToLong(BatchStats::getWritingTextDuration)
|
||||
.toArray();
|
||||
}
|
||||
|
||||
|
||||
@ -1,9 +1,8 @@
|
||||
package com.knecon.fforesight.service.ocr.processor.service;
|
||||
|
||||
import java.util.List;
|
||||
import java.util.Map;
|
||||
import java.util.Set;
|
||||
import java.util.concurrent.CompletableFuture;
|
||||
import java.util.function.Supplier;
|
||||
|
||||
import org.slf4j.MDC;
|
||||
import org.springframework.stereotype.Service;
|
||||
@ -23,7 +22,6 @@ import com.pdftron.pdf.PDFDoc;
|
||||
|
||||
import lombok.AccessLevel;
|
||||
import lombok.RequiredArgsConstructor;
|
||||
import lombok.SneakyThrows;
|
||||
import lombok.experimental.FieldDefaults;
|
||||
import lombok.extern.slf4j.Slf4j;
|
||||
import reactor.core.publisher.Mono;
|
||||
@ -69,10 +67,6 @@ public class AsyncOcrService {
|
||||
|
||||
private void beginAnalysis(BinaryData data, BatchContext batchContext, Set<AzureOcrFeature> features) throws InterruptedException {
|
||||
|
||||
if (settings.isUseCaches() && batchContext.batch().getAzureResultCacheFile().exists()) {
|
||||
handleCached(batchContext);
|
||||
}
|
||||
|
||||
batchContext.supervisor.enterConcurrency(batchContext.batch);
|
||||
|
||||
batchContext.supervisor.logUploadStart(batchContext.batch, data.getLength());
|
||||
@ -91,55 +85,33 @@ public class AsyncOcrService {
|
||||
if (LongRunningOperationStatus.SUCCESSFULLY_COMPLETED == response.getStatus()) {
|
||||
return response.getFinalResult();
|
||||
}
|
||||
String message = "Polling completed unsuccessfully with status: " + response.getStatus();
|
||||
log.error(message);
|
||||
return Mono.error(new IllegalStateException(message));
|
||||
return Mono.error(new IllegalStateException("Polling completed unsuccessfully with status: " + response.getStatus()));
|
||||
}).subscribe(finalResult -> handleSuccessful(finalResult, batchContext),//
|
||||
ex -> handleError(ex, batchContext),//
|
||||
() -> handleCompleted(batchContext));
|
||||
}
|
||||
|
||||
|
||||
@SneakyThrows
|
||||
private static void handleCached(BatchContext batchContext) {
|
||||
|
||||
var mdcContext = MDC.getCopyOfContextMap();
|
||||
Thread thread = new Thread(() -> {
|
||||
MDC.setContextMap(mdcContext);
|
||||
log.info("Batch {}: Using cached ocr result", batchContext.batch.getIndex());
|
||||
batchContext.batchStats().finishUpload();
|
||||
batchContext.batchStats().finishApiWait();
|
||||
batchContext.supervisor.logPageSuccess(batchContext.batch());
|
||||
try {
|
||||
batchContext.layerFactory.processAnalyzeResult(batchContext.batch(), batchContext.batch().getAzureResultCache());
|
||||
} catch (InterruptedException e) {
|
||||
batchContext.supervisor.logPageError(batchContext.batch, e);
|
||||
}
|
||||
|
||||
});
|
||||
thread.start();
|
||||
}
|
||||
|
||||
|
||||
private static void handleCompleted(BatchContext batchContext) {
|
||||
|
||||
log.info("Batch {}: Completed with pages {}", batchContext.batch.getIndex(), batchContext.batch);
|
||||
log.info("Completed : {}", batchContext.batch);
|
||||
batchContext.supervisor.leaveConcurrency(batchContext.batch);
|
||||
}
|
||||
|
||||
|
||||
private void handleError(Throwable ex, BatchContext batchContext) {
|
||||
|
||||
batchContext.supervisor.leaveConcurrency(batchContext.batch);
|
||||
batchContext.supervisor.logPageError(batchContext.batch, ex);
|
||||
}
|
||||
|
||||
|
||||
private void handleSuccessful(AnalyzeResult finalResult, BatchContext batchContext) {
|
||||
|
||||
batchContext.supervisor.leaveConcurrency(batchContext.batch);
|
||||
try {
|
||||
mapper.writeValue(batchContext.batch().getAzureResultCacheFile(), finalResult);
|
||||
batchContext.supervisor.logPageSuccess(batchContext.batch());
|
||||
if (settings.isDebug()) {
|
||||
mapper.writeValue(batchContext.batch().getImagePipelineDir().resolve("azure_result_%d.json".formatted(batchContext.batch().getIndex())).toFile(), finalResult);
|
||||
}
|
||||
batchContext.layerFactory.processAnalyzeResult(batchContext.batch(), finalResult);
|
||||
} catch (Exception e) {
|
||||
handleError(e, batchContext);
|
||||
|
||||
@ -6,14 +6,15 @@ import java.util.Set;
|
||||
|
||||
import org.springframework.beans.factory.annotation.Value;
|
||||
import org.springframework.stereotype.Service;
|
||||
import org.yaml.snakeyaml.events.Event;
|
||||
|
||||
import com.azure.ai.documentintelligence.DocumentIntelligenceAsyncClient;
|
||||
import com.azure.ai.documentintelligence.DocumentIntelligenceClientBuilder;
|
||||
import com.azure.ai.documentintelligence.models.AnalyzeDocumentOptions;
|
||||
import com.azure.ai.documentintelligence.models.AnalyzeOperationDetails;
|
||||
import com.azure.ai.documentintelligence.models.AnalyzeDocumentRequest;
|
||||
import com.azure.ai.documentintelligence.models.AnalyzeResult;
|
||||
import com.azure.ai.documentintelligence.models.AnalyzeResultOperation;
|
||||
import com.azure.ai.documentintelligence.models.ContentFormat;
|
||||
import com.azure.ai.documentintelligence.models.DocumentAnalysisFeature;
|
||||
import com.azure.ai.documentintelligence.models.DocumentContentFormat;
|
||||
import com.azure.ai.documentintelligence.models.StringIndexType;
|
||||
import com.azure.core.credential.AzureKeyCredential;
|
||||
import com.azure.core.util.BinaryData;
|
||||
@ -44,23 +45,28 @@ public class AzureOcrResource {
|
||||
|
||||
|
||||
@SneakyThrows
|
||||
public PollerFlux<AnalyzeOperationDetails, AnalyzeResult> callAzureAsync(BinaryData data, Set<AzureOcrFeature> features) {
|
||||
public PollerFlux<AnalyzeResultOperation, AnalyzeResult> callAzureAsync(BinaryData data, Set<AzureOcrFeature> features) {
|
||||
|
||||
AnalyzeDocumentOptions analyzeDocumentOptions = new AnalyzeDocumentOptions(data.toBytes());
|
||||
analyzeDocumentOptions.setStringIndexType(StringIndexType.UTF16_CODE_UNIT);
|
||||
analyzeDocumentOptions.setDocumentAnalysisFeatures(buildFeatures(features));
|
||||
analyzeDocumentOptions.setOutputContentFormat(buildContentFormat());
|
||||
return asyncClient.beginAnalyzeDocument(getModelId(features), analyzeDocumentOptions);
|
||||
AnalyzeDocumentRequest analyzeRequest = new AnalyzeDocumentRequest().setBase64Source(data.toBytes());
|
||||
|
||||
return asyncClient.beginAnalyzeDocument(getModelId(features),
|
||||
null,
|
||||
null,
|
||||
StringIndexType.UTF16CODE_UNIT,
|
||||
buildFeatures(features),
|
||||
null,
|
||||
buildContentFormat(),
|
||||
analyzeRequest);
|
||||
|
||||
}
|
||||
|
||||
|
||||
private DocumentContentFormat buildContentFormat() {
|
||||
private ContentFormat buildContentFormat() {
|
||||
|
||||
if (Objects.equal(settings.getContentFormat(), "markdown")) {
|
||||
return DocumentContentFormat.MARKDOWN;
|
||||
return ContentFormat.MARKDOWN;
|
||||
}
|
||||
return DocumentContentFormat.TEXT;
|
||||
return ContentFormat.TEXT;
|
||||
}
|
||||
|
||||
|
||||
@ -80,9 +86,6 @@ public class AzureOcrResource {
|
||||
if (features.contains(AzureOcrFeature.IDP)) {
|
||||
azureFeatures.add(DocumentAnalysisFeature.KEY_VALUE_PAIRS);
|
||||
}
|
||||
if (settings.isAzureFontStyleDetection() && features.contains(AzureOcrFeature.FONT_STYLE_DETECTION)) {
|
||||
azureFeatures.add(DocumentAnalysisFeature.STYLE_FONT);
|
||||
}
|
||||
azureFeatures.add(DocumentAnalysisFeature.BARCODES);
|
||||
|
||||
return azureFeatures;
|
||||
|
||||
@ -1,12 +1,7 @@
|
||||
package com.knecon.fforesight.service.ocr.processor.service;
|
||||
|
||||
import static com.knecon.fforesight.service.ocr.processor.utils.ListSplittingUtils.formatIntervals;
|
||||
|
||||
import java.nio.file.Files;
|
||||
import java.nio.file.Path;
|
||||
import java.util.ArrayList;
|
||||
import java.util.Collections;
|
||||
import java.util.HashSet;
|
||||
import java.util.List;
|
||||
import java.util.Set;
|
||||
|
||||
@ -23,10 +18,9 @@ import com.pdftron.sdf.SDFDoc;
|
||||
import lombok.AccessLevel;
|
||||
import lombok.RequiredArgsConstructor;
|
||||
import lombok.SneakyThrows;
|
||||
import lombok.experimental.FieldDefaults;
|
||||
import lombok.extern.slf4j.Slf4j;
|
||||
|
||||
@Slf4j
|
||||
import lombok.experimental.FieldDefaults;
|
||||
|
||||
@Service
|
||||
@RequiredArgsConstructor
|
||||
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
|
||||
@ -35,39 +29,26 @@ public class BatchFactory {
|
||||
OcrServiceSettings settings;
|
||||
|
||||
|
||||
@SneakyThrows
|
||||
public List<PageBatch> splitIntoBatches(PDFDoc pdfDoc, OcrExecutionSupervisor supervisor, Set<AzureOcrFeature> features, Path runDir) {
|
||||
public static String formatBatchFilename(int number) {
|
||||
|
||||
Set<Integer> pagesToProcess = findPagesToProcess(pdfDoc, features);
|
||||
supervisor.logImageExtractionFinished(pdfDoc.getPageCount(), pagesToProcess.size());
|
||||
|
||||
List<PageBatch> batches = buildBatches(pdfDoc, supervisor, features, runDir, pagesToProcess);
|
||||
if (batches.size() > 1) {
|
||||
log.info("Split {} pages to process into {} batches", pagesToProcess.size(), batches.size());
|
||||
}
|
||||
return batches;
|
||||
return "batch_%d.pdf".formatted(number);
|
||||
}
|
||||
|
||||
|
||||
@SneakyThrows
|
||||
public Set<Integer> findPagesToProcess(PDFDoc pdfDoc, Set<AzureOcrFeature> features) {
|
||||
public List<PageBatch> splitIntoBatches(PDFDoc pdfDoc, OcrExecutionSupervisor supervisor, Set<AzureOcrFeature> features, Path fileDir) {
|
||||
|
||||
if (features.contains(AzureOcrFeature.ALL_PAGES)) {
|
||||
Set<Integer> pages = new HashSet<>();
|
||||
for (int i = 1; i <= pdfDoc.getPageCount(); i++) {
|
||||
pages.add(i);
|
||||
}
|
||||
return Collections.unmodifiableSet(pages);
|
||||
}
|
||||
Set<Integer> pagesWithImages = ImageDetectionService.findPagesToProcess(pdfDoc, features);
|
||||
supervisor.logImageExtractionFinished(pdfDoc.getPageCount(), pagesWithImages.size());
|
||||
|
||||
return ImageDetectionService.findPagesWithImages(pdfDoc);
|
||||
return buildBatches(pdfDoc, supervisor, features, fileDir, pagesWithImages);
|
||||
}
|
||||
|
||||
|
||||
public List<PageBatch> buildBatches(PDFDoc pdfDoc,
|
||||
OcrExecutionSupervisor supervisor,
|
||||
Set<AzureOcrFeature> features,
|
||||
Path runDir,
|
||||
Path fileDir,
|
||||
Set<Integer> pagesWithImages) throws PDFNetException {
|
||||
|
||||
List<PageBatch> batches = new ArrayList<>();
|
||||
@ -79,48 +60,30 @@ public class BatchFactory {
|
||||
}
|
||||
numbersForCurrentBatch.add(pageNumber);
|
||||
if (numbersForCurrentBatch.size() == settings.getBatchSize()) {
|
||||
batches.add(create(batches.size(), pdfDoc, numbersForCurrentBatch, runDir));
|
||||
batches.add(create(batches.size(), pdfDoc, numbersForCurrentBatch, fileDir));
|
||||
numbersForCurrentBatch = new ArrayList<>();
|
||||
}
|
||||
}
|
||||
if (!numbersForCurrentBatch.isEmpty()) {
|
||||
batches.add(create(batches.size(), pdfDoc, numbersForCurrentBatch, runDir));
|
||||
batches.add(create(batches.size(), pdfDoc, numbersForCurrentBatch, fileDir));
|
||||
}
|
||||
return batches;
|
||||
}
|
||||
|
||||
|
||||
@SneakyThrows
|
||||
public static PageBatch create(int number, PDFDoc pdfDoc, List<Integer> pageNumbers, Path runDir) {
|
||||
public static PageBatch create(int number, PDFDoc pdfDoc, List<Integer> pageNumbers, Path fileDir) {
|
||||
|
||||
if (pageNumbers.isEmpty()) {
|
||||
throw new IllegalArgumentException("pageNumbers must not be empty");
|
||||
}
|
||||
Path batchDir = formatBatchDir(number, pageNumbers, runDir);
|
||||
Files.createDirectories(batchDir);
|
||||
|
||||
Path batchDocPath = batchDir.resolve("batch.pdf");
|
||||
Path batchDocPath = fileDir.resolve(formatBatchFilename(number));
|
||||
try (var batchDoc = extractBatchDocument(pdfDoc, pageNumbers)) {
|
||||
Optimizer.optimize(batchDoc);
|
||||
batchDoc.save(batchDocPath.toFile().toString(), SDFDoc.SaveMode.LINEARIZED, null);
|
||||
}
|
||||
PageBatch batch = new PageBatch(number, pageNumbers, batchDocPath, batchDir);
|
||||
Files.createDirectories(batch.getRenderedImageDir());
|
||||
Files.createDirectories(batch.getProcessedImageDir());
|
||||
return batch;
|
||||
}
|
||||
|
||||
|
||||
private static Path formatBatchDir(int number, List<Integer> pageNumbers, Path runDir) {
|
||||
|
||||
List<String> intervals = formatIntervals(pageNumbers);
|
||||
if (intervals.size() > 4) {
|
||||
intervals = intervals.subList(0, 4);
|
||||
intervals.add("...");
|
||||
}
|
||||
|
||||
String batchName = String.join(", ", intervals);
|
||||
return runDir.resolve("batch_%04d_%s".formatted(number, batchName));
|
||||
return new PageBatch(number, pageNumbers, batchDocPath, fileDir);
|
||||
}
|
||||
|
||||
|
||||
|
||||
@ -10,7 +10,7 @@ public class BatchStats {
|
||||
|
||||
private long apiWaitTimestamp = -1;
|
||||
private long imageUploadTimestamp = -1;
|
||||
private long mappingResultTimestamp = -1;
|
||||
private long writingTextTimestamp = -1;
|
||||
private long batchRenderTimestamp = -1;
|
||||
|
||||
|
||||
@ -38,9 +38,9 @@ public class BatchStats {
|
||||
}
|
||||
|
||||
|
||||
public void finishMappingResult() {
|
||||
public void finishWritingText() {
|
||||
|
||||
mappingResultTimestamp = System.currentTimeMillis();
|
||||
writingTextTimestamp = System.currentTimeMillis();
|
||||
}
|
||||
|
||||
|
||||
@ -50,33 +50,15 @@ public class BatchStats {
|
||||
}
|
||||
|
||||
|
||||
public boolean isApiWaitFinished() {
|
||||
|
||||
return apiWaitTimestamp > 0;
|
||||
}
|
||||
|
||||
|
||||
public boolean isMappingResultFinished() {
|
||||
|
||||
return mappingResultTimestamp > 0;
|
||||
}
|
||||
|
||||
|
||||
public boolean isBatchRenderFinished() {
|
||||
|
||||
return batchRenderTimestamp > 0;
|
||||
}
|
||||
|
||||
|
||||
public long getApiWaitDuration() {return this.apiWaitTimestamp - imageUploadTimestamp;}
|
||||
|
||||
|
||||
public long getImageUploadDuration() {return this.imageUploadTimestamp - batchRenderTimestamp;}
|
||||
|
||||
|
||||
public long getMappingResultDuration() {return this.mappingResultTimestamp - apiWaitTimestamp;}
|
||||
public long getWritingTextDuration() {return this.writingTextTimestamp - apiWaitTimestamp;}
|
||||
|
||||
|
||||
public long getBatchRenderDuration() {return startTimestamp - this.batchRenderTimestamp;}
|
||||
public long getBatchRenderDuration() {return this.batchRenderTimestamp - startTimestamp;}
|
||||
|
||||
}
|
||||
|
||||
@ -1,4 +1,4 @@
|
||||
package com.knecon.fforesight.service.ocr.v1.server;
|
||||
package com.knecon.fforesight.service.ocr.processor.service;
|
||||
|
||||
import java.io.File;
|
||||
import java.io.FileInputStream;
|
||||
@ -33,23 +33,20 @@ public class FileStorageService {
|
||||
public void storeFiles(DocumentRequest request, File documentFile, File viewerDocumentFile, File analyzeResultFile) {
|
||||
|
||||
try (var in = new FileInputStream(viewerDocumentFile)) {
|
||||
if (request.optionalViewerDocumentId()
|
||||
.isPresent()) {
|
||||
if (request.optionalViewerDocumentId().isPresent()) {
|
||||
storageService.storeObject(TenantContext.getTenantId(), request.getViewerDocId(), in);
|
||||
} else {
|
||||
storageService.storeObject(TenantContext.getTenantId(), getStorageId(request.getDossierId(), request.getFileId(), FileType.VIEWER_DOCUMENT), in);
|
||||
}
|
||||
}
|
||||
try (var in = new FileInputStream(documentFile)) {
|
||||
if (request.optionalOriginDocumentId()
|
||||
.isPresent()) {
|
||||
if (request.optionalOriginDocumentId().isPresent()) {
|
||||
storageService.storeObject(TenantContext.getTenantId(), request.getOriginDocumentId(), in);
|
||||
} else {
|
||||
storageService.storeObject(TenantContext.getTenantId(), getStorageId(request.getDossierId(), request.getFileId(), FileType.ORIGIN), in);
|
||||
}
|
||||
}
|
||||
if (request.optionalIdpResultId()
|
||||
.isPresent() && analyzeResultFile.exists()) {
|
||||
if (request.optionalIdpResultId().isPresent()) {
|
||||
try (var in = new FileInputStream(analyzeResultFile)) {
|
||||
storageService.storeObject(TenantContext.getTenantId(), request.getIdpResultId(), in);
|
||||
}
|
||||
@ -62,8 +59,7 @@ public class FileStorageService {
|
||||
|
||||
Files.createDirectories(documentFile.getParentFile().toPath());
|
||||
|
||||
String originDocumentId = request.optionalOriginDocumentId()
|
||||
.orElse(getStorageId(request.getDossierId(), request.getFileId(), FileType.ORIGIN));
|
||||
String originDocumentId = request.optionalOriginDocumentId().orElse(getStorageId(request.getDossierId(), request.getFileId(), FileType.ORIGIN));
|
||||
|
||||
storageService.downloadTo(TenantContext.getTenantId(), originDocumentId, documentFile);
|
||||
|
||||
@ -1,24 +1,16 @@
|
||||
package com.knecon.fforesight.service.ocr.processor.service;
|
||||
|
||||
import java.util.Set;
|
||||
|
||||
import org.springframework.stereotype.Service;
|
||||
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.AzureOcrFeature;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.DocumentRequest;
|
||||
|
||||
@Service
|
||||
public interface IOcrMessageSender {
|
||||
|
||||
void sendUpdate(String fileId, int finishedImages, int totalImages, Set<AzureOcrFeature> features);
|
||||
void sendUpdate(String fileId, int finishedImages, int totalImages);
|
||||
|
||||
void sendOCRStarted(String fileId);
|
||||
|
||||
void sendOCRStarted(String fileId, Set<AzureOcrFeature> features);
|
||||
void sendOcrFinished(String fileId, int totalImages);
|
||||
|
||||
|
||||
void sendOcrFinished(String fileId, int totalImages, Set<AzureOcrFeature> features);
|
||||
|
||||
|
||||
void sendOcrResponse(DocumentRequest request);
|
||||
void sendOcrResponse(String dossierId, String fileId);
|
||||
|
||||
}
|
||||
|
||||
@ -24,7 +24,21 @@ public class ImageDetectionService {
|
||||
|
||||
|
||||
@SneakyThrows
|
||||
public Set<Integer> findPagesWithImages(PDFDoc pdfDoc) {
|
||||
public Set<Integer> findPagesToProcess(PDFDoc pdfDoc, Set<AzureOcrFeature> features) {
|
||||
|
||||
if (features.contains(AzureOcrFeature.ALL_PAGES)) {
|
||||
Set<Integer> pages = new HashSet<>();
|
||||
for (int i = 1; i <= pdfDoc.getPageCount(); i++) {
|
||||
pages.add(i);
|
||||
}
|
||||
return Collections.unmodifiableSet(pages);
|
||||
}
|
||||
|
||||
return findPagesWithImages(pdfDoc);
|
||||
}
|
||||
|
||||
|
||||
private Set<Integer> findPagesWithImages(PDFDoc pdfDoc) throws PDFNetException {
|
||||
|
||||
Set<Integer> pagesWithImages = new HashSet<>();
|
||||
try (ElementReader reader = new ElementReader()) {
|
||||
|
||||
@ -5,7 +5,6 @@ import static com.knecon.fforesight.service.ocr.processor.model.Statistics.human
|
||||
import java.io.File;
|
||||
import java.io.FileInputStream;
|
||||
import java.io.FileOutputStream;
|
||||
import java.io.IOException;
|
||||
import java.nio.file.Files;
|
||||
import java.nio.file.Path;
|
||||
import java.nio.file.StandardCopyOption;
|
||||
@ -14,7 +13,6 @@ import java.util.Set;
|
||||
|
||||
import org.springframework.stereotype.Service;
|
||||
|
||||
import com.fasterxml.jackson.databind.ObjectMapper;
|
||||
import com.iqser.red.pdftronlogic.commons.InvisibleElementRemovalService;
|
||||
import com.iqser.red.pdftronlogic.commons.OCGWatermarkRemovalService;
|
||||
import com.iqser.red.pdftronlogic.commons.WatermarkRemovalService;
|
||||
@ -40,6 +38,8 @@ import lombok.extern.slf4j.Slf4j;
|
||||
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
|
||||
public class OCRService {
|
||||
|
||||
public static final String IMAGE_PIPELINE_DIR = "image_pipeline";
|
||||
public static final String AZURE_OUTPUT_DIR = "azure_output";
|
||||
IOcrMessageSender ocrMessageSender;
|
||||
WatermarkRemovalService watermarkRemovalService;
|
||||
InvisibleElementRemovalService invisibleElementRemovalService;
|
||||
@ -47,7 +47,6 @@ public class OCRService {
|
||||
BatchFactory batchFactory;
|
||||
AsyncOcrService asyncOcrService;
|
||||
OcrServiceSettings settings;
|
||||
ObjectMapper mapper;
|
||||
|
||||
|
||||
/**
|
||||
@ -60,10 +59,10 @@ public class OCRService {
|
||||
* @param tmpDir working directory for all files
|
||||
* @param documentFile the file to perform ocr on, results are written invisibly
|
||||
* @param viewerDocumentFile debugging file, results are written visibly in an optional content group
|
||||
* @param idpResultFile result file with additional information
|
||||
* @param analyzeResultFile result file with additional information
|
||||
*/
|
||||
@Observed(name = "OCRService", contextualName = "run-ocr-on-document")
|
||||
public void runOcrOnDocument(String dossierId, String fileId, Set<AzureOcrFeature> features, Path tmpDir, File documentFile, File viewerDocumentFile, File idpResultFile) {
|
||||
public void runOcrOnDocument(String dossierId, String fileId, Set<AzureOcrFeature> features, Path tmpDir, File documentFile, File viewerDocumentFile, File analyzeResultFile) {
|
||||
|
||||
if (features.contains(AzureOcrFeature.REMOVE_WATERMARKS)) {
|
||||
removeWatermark(documentFile);
|
||||
@ -71,9 +70,10 @@ public class OCRService {
|
||||
|
||||
removeInvisibleElements(documentFile);
|
||||
|
||||
log.info("Starting OCR");
|
||||
long ocrStart = System.currentTimeMillis();
|
||||
|
||||
Statistics stats = runOcr(tmpDir, documentFile, viewerDocumentFile, fileId, dossierId, idpResultFile, features).getStatistics();
|
||||
Statistics stats = runOcr(tmpDir, documentFile, viewerDocumentFile, fileId, dossierId, analyzeResultFile, features).getStatistics();
|
||||
|
||||
long ocrEnd = System.currentTimeMillis();
|
||||
log.info("OCR successful, took {}", humanizeDuration(ocrEnd - ocrStart));
|
||||
@ -120,22 +120,28 @@ public class OCRService {
|
||||
|
||||
|
||||
@SneakyThrows
|
||||
public OcrExecutionSupervisor runOcr(Path runDir,
|
||||
public OcrExecutionSupervisor runOcr(Path tmpDir,
|
||||
File documentFile,
|
||||
File viewerDocumentFile,
|
||||
String fileId,
|
||||
String dossierId,
|
||||
File idpResultFile,
|
||||
File analyzeResultFile,
|
||||
Set<AzureOcrFeature> features) {
|
||||
|
||||
Path tmpImageDir = tmpDir.resolve(IMAGE_PIPELINE_DIR);
|
||||
Path azureOutputDir = tmpDir.resolve(AZURE_OUTPUT_DIR);
|
||||
|
||||
Files.createDirectories(azureOutputDir);
|
||||
Files.createDirectories(tmpImageDir);
|
||||
|
||||
try (var in = new FileInputStream(documentFile); PDFDoc pdfDoc = new PDFDoc(in)) {
|
||||
|
||||
OCGWatermarkRemovalService.removeWatermarks(pdfDoc);
|
||||
|
||||
OcrExecutionSupervisor supervisor = new OcrExecutionSupervisor(pdfDoc.getPageCount(), ocrMessageSender, fileId, settings, features);
|
||||
OcrExecutionSupervisor supervisor = new OcrExecutionSupervisor(pdfDoc.getPageCount(), ocrMessageSender, fileId, settings);
|
||||
supervisor.getStatistics().setStart();
|
||||
|
||||
List<PageBatch> batches = batchFactory.splitIntoBatches(pdfDoc, supervisor, features, runDir);
|
||||
List<PageBatch> batches = batchFactory.splitIntoBatches(pdfDoc, supervisor, features, tmpImageDir);
|
||||
|
||||
OcrResult ocrResult = asyncOcrService.awaitOcr(pdfDoc, supervisor, features, batches);
|
||||
|
||||
@ -147,10 +153,6 @@ public class OCRService {
|
||||
RotationCorrectionUtility.rotatePages(viewerDocumentFile.toPath(), viewerDocumentFile.toPath(), ocrResult.anglesPerPage());
|
||||
}
|
||||
|
||||
if (features.contains(AzureOcrFeature.IDP)) {
|
||||
saveIdpResultFile(idpResultFile, ocrResult);
|
||||
}
|
||||
|
||||
supervisor.getStatistics().drawingPdfFinished();
|
||||
|
||||
supervisor.sendFinished();
|
||||
@ -160,12 +162,4 @@ public class OCRService {
|
||||
|
||||
}
|
||||
|
||||
|
||||
private void saveIdpResultFile(File idpResultFile, OcrResult ocrResult) throws IOException {
|
||||
|
||||
try (var out = new FileOutputStream(idpResultFile)) {
|
||||
mapper.writeValue(out, ocrResult.idpResult());
|
||||
}
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
@ -5,7 +5,6 @@ import static com.knecon.fforesight.service.ocr.processor.model.Statistics.human
|
||||
|
||||
import java.util.Collections;
|
||||
import java.util.HashSet;
|
||||
import java.util.Map;
|
||||
import java.util.Set;
|
||||
import java.util.concurrent.ArrayBlockingQueue;
|
||||
import java.util.concurrent.BlockingQueue;
|
||||
@ -14,7 +13,6 @@ import java.util.concurrent.CountDownLatch;
|
||||
import com.knecon.fforesight.service.ocr.processor.OcrServiceSettings;
|
||||
import com.knecon.fforesight.service.ocr.processor.model.PageBatch;
|
||||
import com.knecon.fforesight.service.ocr.processor.model.Statistics;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.AzureOcrFeature;
|
||||
|
||||
import lombok.AccessLevel;
|
||||
import lombok.Getter;
|
||||
@ -40,15 +38,12 @@ public class OcrExecutionSupervisor {
|
||||
|
||||
String fileId;
|
||||
|
||||
Set<AzureOcrFeature> features;
|
||||
|
||||
|
||||
public OcrExecutionSupervisor(int totalPageCount, IOcrMessageSender ocrMessageSender, String fileId, OcrServiceSettings settings, Set<AzureOcrFeature> features) {
|
||||
public OcrExecutionSupervisor(int totalPageCount, IOcrMessageSender ocrMessageSender, String fileId, OcrServiceSettings settings) {
|
||||
|
||||
this.totalPageCount = totalPageCount;
|
||||
this.ocrMessageSender = ocrMessageSender;
|
||||
this.fileId = fileId;
|
||||
this.features = features;
|
||||
this.errorPages = Collections.synchronizedSet(new HashSet<>());
|
||||
this.countDownPagesToProcess = new CountDownLatch(totalPageCount);
|
||||
this.statistics = new Statistics();
|
||||
@ -70,16 +65,16 @@ public class OcrExecutionSupervisor {
|
||||
}
|
||||
|
||||
|
||||
public void logImageExtractionFinished(int numberOfPages, int numberOfPagesToProcess) {
|
||||
public void logImageExtractionFinished(int numberOfPages, int numberOfImages) {
|
||||
|
||||
statistics.imageExtractionFinished();
|
||||
log.info("Images found on {}/{} pages in {}", numberOfPagesToProcess, numberOfPages, humanizeDuration(statistics.getImageExtractionDuration()));
|
||||
log.info("Images found on {}/{} pages in {}", numberOfImages, numberOfPages, humanizeDuration(statistics.getImageExtractionDuration()));
|
||||
}
|
||||
|
||||
|
||||
public void logUploadStart(PageBatch pageRange, long bytes) {
|
||||
|
||||
log.info("Batch {}: Start uploading pages {} with {}", pageRange.getIndex(), pageRange, humanizeBytes(bytes));
|
||||
log.info("Start uploading pages {} with {}", pageRange, humanizeBytes(bytes));
|
||||
statistics.getBatchStats(pageRange).start();
|
||||
statistics.increaseTotalBytes(pageRange, bytes);
|
||||
}
|
||||
@ -88,11 +83,11 @@ public class OcrExecutionSupervisor {
|
||||
public void logInProgress(PageBatch pageRange) {
|
||||
|
||||
if (!statistics.getBatchStats(pageRange).isUploadFinished()) {
|
||||
log.info("Batch {}: Pages {} is in progress", pageRange.getIndex(), pageRange);
|
||||
log.info("Pages {} is in progress", pageRange);
|
||||
statistics.getBatchStats(pageRange).finishUpload();
|
||||
ocrMessageSender.sendUpdate(fileId, processedPages(), getTotalPageCount(), features);
|
||||
ocrMessageSender.sendUpdate(fileId, processedPages(), getTotalPageCount());
|
||||
} else {
|
||||
log.debug("Batch {}: Pages {} still in progress", pageRange.getIndex(), pageRange);
|
||||
log.debug("Pages {} still in progress", pageRange);
|
||||
}
|
||||
}
|
||||
|
||||
@ -100,16 +95,15 @@ public class OcrExecutionSupervisor {
|
||||
public void finishMappingResult(PageBatch batch) {
|
||||
|
||||
batch.forEach(pageIndex -> countDownPagesToProcess.countDown());
|
||||
statistics.getBatchStats(batch).finishMappingResult();
|
||||
ocrMessageSender.sendUpdate(fileId, this.processedPages(), getTotalPageCount(), features);
|
||||
log.info("Batch {}: Finished mapping result with pages {}", batch.getIndex(), batch);
|
||||
statistics.getBatchStats(batch).finishWritingText();
|
||||
ocrMessageSender.sendUpdate(fileId, this.processedPages(), getTotalPageCount());
|
||||
}
|
||||
|
||||
|
||||
public void logPageSkipped(Integer pageIndex) {
|
||||
|
||||
this.countDownPagesToProcess.countDown();
|
||||
ocrMessageSender.sendUpdate(fileId, this.processedPages(), getTotalPageCount(), features);
|
||||
ocrMessageSender.sendUpdate(fileId, this.processedPages(), getTotalPageCount());
|
||||
log.debug("{}/{}: No images to ocr on page {}", processedPages(), getTotalPageCount(), pageIndex);
|
||||
|
||||
}
|
||||
@ -119,43 +113,21 @@ public class OcrExecutionSupervisor {
|
||||
|
||||
this.errorPages.add(batch);
|
||||
batch.forEach(pageIndex -> this.countDownPagesToProcess.countDown());
|
||||
ocrMessageSender.sendUpdate(fileId, this.processedPages(), getTotalPageCount(), features);
|
||||
log.error("{}/{}: Error occurred in batch {} with pages {}", processedPages(), getTotalPageCount(), batch.getIndex(), batch, e);
|
||||
ocrMessageSender.sendUpdate(fileId, this.processedPages(), getTotalPageCount());
|
||||
log.error("{}/{}: Error occurred on pages {}", processedPages(), getTotalPageCount(), batch, e);
|
||||
}
|
||||
|
||||
|
||||
public void logPageSuccess(PageBatch batch) {
|
||||
|
||||
statistics.getBatchStats(batch).finishApiWait();
|
||||
log.info("{}/{}: Finished OCR in batch {} with pages {}", processedPages(), getTotalPageCount(), batch.getIndex(), batch);
|
||||
log.info("{}/{}: Finished OCR on pages {}", processedPages(), getTotalPageCount(), batch);
|
||||
}
|
||||
|
||||
|
||||
private int processedPages() {
|
||||
|
||||
if (countDownPagesToProcess.getCount() == 0) {
|
||||
return totalPageCount;
|
||||
}
|
||||
int processedPages = 0;
|
||||
for (Map.Entry<PageBatch, BatchStats> entry : statistics.getBatchStats().entrySet()) {
|
||||
PageBatch pageBatch = entry.getKey();
|
||||
BatchStats batchStats = entry.getValue();
|
||||
float percentage = 0;
|
||||
if (batchStats.isBatchRenderFinished()) {
|
||||
percentage += 0.1f;
|
||||
}
|
||||
if (batchStats.isUploadFinished()) {
|
||||
percentage += 0.3f;
|
||||
}
|
||||
if (batchStats.isApiWaitFinished()) {
|
||||
percentage += 0.3f;
|
||||
}
|
||||
if (batchStats.isMappingResultFinished()) {
|
||||
percentage += 0.3f;
|
||||
}
|
||||
processedPages += (int) (pageBatch.size() * percentage);
|
||||
}
|
||||
return processedPages;
|
||||
return (int) (totalPageCount - countDownPagesToProcess.getCount());
|
||||
}
|
||||
|
||||
|
||||
@ -172,7 +144,7 @@ public class OcrExecutionSupervisor {
|
||||
requireNoErrors();
|
||||
|
||||
log.info("{}/{}: Finished OCR on all pages", getTotalPageCount(), getTotalPageCount());
|
||||
ocrMessageSender.sendOcrFinished(fileId, getTotalPageCount(), features);
|
||||
ocrMessageSender.sendOcrFinished(fileId, getTotalPageCount());
|
||||
}
|
||||
|
||||
|
||||
|
||||
@ -7,13 +7,12 @@ import java.util.Optional;
|
||||
|
||||
import com.azure.ai.documentintelligence.models.DocumentPage;
|
||||
import com.azure.ai.documentintelligence.models.DocumentWord;
|
||||
import com.knecon.fforesight.service.ocr.processor.service.OcrResultPostProcessingPipeline;
|
||||
import com.knecon.fforesight.service.ocr.processor.visualizations.WritableOcrResultFactory;
|
||||
import com.knecon.fforesight.service.ocr.processor.visualizations.utils.RotationCorrectionUtility;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.QuadPoint;
|
||||
import com.sun.jna.Pointer;
|
||||
|
||||
import lombok.SneakyThrows;
|
||||
import lombok.experimental.UtilityClass;
|
||||
import lombok.extern.slf4j.Slf4j;
|
||||
import net.sourceforge.lept4j.Leptonica1;
|
||||
import net.sourceforge.lept4j.Numa;
|
||||
@ -24,35 +23,26 @@ import net.sourceforge.lept4j.util.LeptUtils;
|
||||
* This class attempts to shrink the BBox of a word to match the exact height of the word. This is only attempted for horizontal or vertical words. Any askew text is left as is.
|
||||
*/
|
||||
@Slf4j
|
||||
@UtilityClass
|
||||
public class BBoxSnuggificationService {
|
||||
|
||||
public static final int PIXEL_COUNT_THRESHOLD = 2; // minimum active pixel count per row for shrinking to stop
|
||||
private static final double AVERAGE_ANGLE_THRESHOLD = 0.2; // Skips snuggification, if the average remaining word rotation of a word, written from left-to-right is bigger than this
|
||||
public static final int INDIVIDUAL_ANGLE_THRESHOLD = 5; // skips snuggification for word, if the remaining rotation is larger than this angle
|
||||
public static final int MAX_SHRINK_PIXELS = 40; // Number of pixels that are allowed to be removed from the top or bottom of an image
|
||||
private static final int MINIMUM_WORD_PIXELS = 5; // Number of pixels that are required for snuggification
|
||||
|
||||
private enum Operation {
|
||||
HORIZONTAL,
|
||||
VERTICAL,
|
||||
BOTH,
|
||||
NONE
|
||||
}
|
||||
|
||||
|
||||
@SneakyThrows
|
||||
public Optional<QuadPoint> snuggify(Pix pageImage, DocumentWord origin, AffineTransform resultToImageTransform) {
|
||||
public static Optional<QuadPoint> snuggify(Pix pageImage, DocumentWord origin, AffineTransform resultToImageTransform) {
|
||||
|
||||
if (pageImage == null) {
|
||||
return Optional.empty();
|
||||
}
|
||||
|
||||
if (origin.getContent().equals("-") || origin.getContent().equals(",")) {
|
||||
// very slim characters should not be snuggified, or the fontsize may be off significantly
|
||||
return Optional.empty();
|
||||
}
|
||||
|
||||
QuadPoint originTransformed = QuadPoint.fromPolygons(origin.getPolygon()).getTransformed(resultToImageTransform);
|
||||
double remainingAngle = Math.abs(RotationCorrectionUtility.getRemainingAngle(originTransformed.getAngle()));
|
||||
QuadPoint.Direction direction = originTransformed.getDirection();
|
||||
@ -63,24 +53,18 @@ public class BBoxSnuggificationService {
|
||||
return Optional.empty();
|
||||
}
|
||||
|
||||
Pix wordImage = OcrResultPostProcessingPipeline.extractWordImage(originTransformed, pageImage);
|
||||
Pix wordImage = WritableOcrResultFactory.extractWordImage(originTransformed, pageImage);
|
||||
|
||||
if (wordImage == null) {
|
||||
log.debug("Unable to extract word image! wordImage: {}, pageImage {}", originTransformed.getBounds2D(), new Rectangle2D.Float(0, 0, pageImage.w, pageImage.h));
|
||||
return Optional.empty();
|
||||
}
|
||||
|
||||
if (!StrokeWidthCalculator.wordImageHasMinimumPixelDensity(wordImage)) {
|
||||
return Optional.empty();
|
||||
}
|
||||
|
||||
Optional<Rectangle2D> snugBox = switch (operation) {
|
||||
case HORIZONTAL -> snuggifyY(wordImage, originTransformed.getBounds2D());
|
||||
case VERTICAL -> snuggifyX(wordImage, originTransformed.getBounds2D());
|
||||
case BOTH -> snuggifyBoth(wordImage, originTransformed);
|
||||
default -> Optional.empty();
|
||||
};
|
||||
|
||||
LeptUtils.disposePix(wordImage);
|
||||
|
||||
AffineTransform imageToResultTransform = resultToImageTransform.createInverse();
|
||||
@ -90,46 +74,31 @@ public class BBoxSnuggificationService {
|
||||
}
|
||||
|
||||
|
||||
private Optional<Rectangle2D> snuggifyBoth(Pix wordImage, QuadPoint originTransformed) {
|
||||
|
||||
Optional<Rectangle2D> snugY = snuggifyY(wordImage, originTransformed.getBounds2D());
|
||||
Optional<Rectangle2D> snugX = snuggifyX(wordImage, originTransformed.getBounds2D());
|
||||
if (snugY.isPresent() && snugX.isPresent()) {
|
||||
return Optional.of(snugY.get().createIntersection(snugX.get()).getBounds2D());
|
||||
} else if (snugY.isPresent()) {
|
||||
return snugY;
|
||||
} else {
|
||||
return snugX;
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
private Operation determineOperation(DocumentWord origin, QuadPoint.Direction direction, double remainingAngle, QuadPoint originTransformed) {
|
||||
private static Operation determineOperation(DocumentWord origin, QuadPoint.Direction direction, double remainingAngle, QuadPoint originTransformed) {
|
||||
|
||||
Operation operation = Operation.NONE;
|
||||
if (((direction.equals(QuadPoint.Direction.RIGHT) || direction.equals(QuadPoint.Direction.LEFT)) && remainingAngle < INDIVIDUAL_ANGLE_THRESHOLD)) {
|
||||
if (((direction.equals(QuadPoint.Direction.RIGHT) || direction.equals(QuadPoint.Direction.LEFT)) && remainingAngle < INDIVIDUAL_ANGLE_THRESHOLD) //
|
||||
|| (origin.getContent().length() < 4 || Math.abs(originTransformed.getAngle()) < AVERAGE_ANGLE_THRESHOLD * 3)) {
|
||||
operation = Operation.HORIZONTAL;
|
||||
} else if ((direction.equals(QuadPoint.Direction.UP) || direction.equals(QuadPoint.Direction.DOWN)) && remainingAngle < INDIVIDUAL_ANGLE_THRESHOLD) {
|
||||
operation = Operation.VERTICAL;
|
||||
} else if ((origin.getContent().length() < 4 || Math.abs(originTransformed.getAngle()) < AVERAGE_ANGLE_THRESHOLD * 3)) {
|
||||
return Operation.BOTH;
|
||||
}
|
||||
return operation;
|
||||
}
|
||||
|
||||
|
||||
private Optional<Rectangle2D> snuggifyX(Pix wordImage, Rectangle2D origin) {
|
||||
private static Optional<Rectangle2D> snuggifyX(Pix wordImage, Rectangle2D origin) {
|
||||
|
||||
Numa colCounts = Leptonica1.pixCountPixelsByColumn(wordImage);
|
||||
int start = 0;
|
||||
int end = wordImage.w - PIXEL_COUNT_THRESHOLD;
|
||||
for (int i = start; i < Math.min(wordImage.w, MAX_SHRINK_PIXELS); i++) {
|
||||
for (int i = start; i < Math.min(wordImage.w, 25); i++) {
|
||||
if (pixCountPerColumn(i, colCounts) > PIXEL_COUNT_THRESHOLD) {
|
||||
start = i;
|
||||
break;
|
||||
}
|
||||
}
|
||||
for (int i = end; i > Math.max(0, wordImage.w - MAX_SHRINK_PIXELS); i--) {
|
||||
for (int i = end; i > Math.max(0, wordImage.w - 25); i--) {
|
||||
if (pixCountPerColumn(i, colCounts) > PIXEL_COUNT_THRESHOLD) {
|
||||
end = i;
|
||||
break;
|
||||
@ -138,24 +107,22 @@ public class BBoxSnuggificationService {
|
||||
if (start == 0 && end == wordImage.w) {
|
||||
return Optional.empty();
|
||||
}
|
||||
if (Math.abs(start - end) < MINIMUM_WORD_PIXELS) {
|
||||
return Optional.empty();
|
||||
}
|
||||
|
||||
return Optional.of(new Rectangle2D.Double(origin.getX() + start, origin.getY(), origin.getWidth() - start - (wordImage.w - end), origin.getHeight()));
|
||||
}
|
||||
|
||||
|
||||
private Optional<Rectangle2D> snuggifyY(Pix wordImage, Rectangle2D origin) {
|
||||
private static Optional<Rectangle2D> snuggifyY(Pix wordImage, Rectangle2D origin) {
|
||||
|
||||
int start = 0;
|
||||
int end = wordImage.h - 1;
|
||||
for (int i = start; i < Math.min(wordImage.h, MAX_SHRINK_PIXELS); i++) {
|
||||
for (int i = start; i < Math.min(wordImage.h, 25); i++) {
|
||||
if (pixCountPerRow(i, wordImage) > PIXEL_COUNT_THRESHOLD) {
|
||||
start = i;
|
||||
break;
|
||||
}
|
||||
}
|
||||
for (int i = end; i > Math.max(0, wordImage.h - MAX_SHRINK_PIXELS); i--) {
|
||||
for (int i = end; i > Math.max(0, wordImage.h - 25); i--) {
|
||||
if (pixCountPerRow(i, wordImage) > PIXEL_COUNT_THRESHOLD) {
|
||||
end = i;
|
||||
break;
|
||||
@ -164,14 +131,11 @@ public class BBoxSnuggificationService {
|
||||
if (start == 0 && end == wordImage.h) {
|
||||
return Optional.empty();
|
||||
}
|
||||
if (Math.abs(start - end) < MINIMUM_WORD_PIXELS) {
|
||||
return Optional.empty();
|
||||
}
|
||||
return Optional.of(new Rectangle2D.Double(origin.getX(), origin.getY() + start, origin.getWidth(), origin.getHeight() - start - (wordImage.h - end)));
|
||||
}
|
||||
|
||||
|
||||
private int pixCountPerRow(int row, Pix pix) {
|
||||
private static int pixCountPerRow(int row, Pix pix) {
|
||||
|
||||
IntBuffer result = IntBuffer.allocate(1);
|
||||
int success = Leptonica1.pixCountPixelsInRow(pix, row, result, null);
|
||||
@ -183,7 +147,7 @@ public class BBoxSnuggificationService {
|
||||
}
|
||||
|
||||
|
||||
private int pixCountPerColumn(int column, Numa colCounts) {
|
||||
private static int pixCountPerColumn(int column, Numa colCounts) {
|
||||
|
||||
if (column > colCounts.n) {
|
||||
throw new IndexOutOfBoundsException("column " + column + " is out of bounds for column count " + colCounts.n);
|
||||
@ -195,7 +159,7 @@ public class BBoxSnuggificationService {
|
||||
}
|
||||
|
||||
|
||||
public boolean canBeSnuggified(DocumentPage resultPage, AffineTransform imageTransform) {
|
||||
public static boolean canBeSnuggified(DocumentPage resultPage, AffineTransform imageTransform) {
|
||||
|
||||
double averageAngle = resultPage.getWords()
|
||||
.stream()
|
||||
|
||||
@ -75,14 +75,13 @@ public class GhostScriptOutputHandler extends Thread {
|
||||
if (line == null) {
|
||||
break;
|
||||
}
|
||||
switch (type) {
|
||||
case STD_OUT -> {
|
||||
log.debug("Batch {}: {}_{}>{}", batchIdx, processName, type.name(), line);
|
||||
addProcessedImageToQueue(line);
|
||||
}
|
||||
case ERROR -> log.error("Batch {}: {}_{}>{}", batchIdx, processName, type.name(), line);
|
||||
}
|
||||
|
||||
if (type.equals(Type.ERROR)) {
|
||||
log.error("{}_{}>{}", processName, type.name(), line);
|
||||
} else {
|
||||
log.debug("{}_{}>{}", processName, type.name(), line);
|
||||
addProcessedImageToQueue(line);
|
||||
}
|
||||
}
|
||||
}
|
||||
is.close();
|
||||
@ -92,7 +91,7 @@ public class GhostScriptOutputHandler extends Thread {
|
||||
if (!pagesToProcess.isEmpty()) {
|
||||
errorHandler.accept(String.format("Ghostscript finished for batch %d, but pages %s remain unprocessed.", batchIdx, formatPagesToProcess()));
|
||||
} else {
|
||||
log.info("Batch {}: rendered successfully!", batchIdx);
|
||||
log.info("{}: Batch rendered successfully!", batchIdx);
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@ -5,95 +5,55 @@ import java.nio.file.Path;
|
||||
import java.util.HashMap;
|
||||
import java.util.List;
|
||||
import java.util.Map;
|
||||
import java.util.concurrent.Semaphore;
|
||||
import java.util.concurrent.TimeUnit;
|
||||
import java.util.function.Consumer;
|
||||
|
||||
import org.springframework.stereotype.Service;
|
||||
|
||||
import com.knecon.fforesight.service.ocr.processor.OcrServiceSettings;
|
||||
import com.knecon.fforesight.service.ocr.processor.model.ImageFile;
|
||||
import com.knecon.fforesight.service.ocr.processor.model.PageBatch;
|
||||
|
||||
import lombok.AccessLevel;
|
||||
import lombok.RequiredArgsConstructor;
|
||||
import lombok.SneakyThrows;
|
||||
import lombok.experimental.FieldDefaults;
|
||||
import lombok.extern.slf4j.Slf4j;
|
||||
|
||||
@Slf4j
|
||||
@Service
|
||||
@RequiredArgsConstructor
|
||||
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
|
||||
@SuppressWarnings("PMD") // can't figure out how to safely close the stdOut and stdError streams in line 72/74
|
||||
public class GhostScriptService {
|
||||
|
||||
private OcrServiceSettings ocrServiceSettings;
|
||||
public static String FORMAT = ".tiff";
|
||||
static String FORMAT = ".tiff";
|
||||
static String DEVICE = "tiffgray";
|
||||
static int DPI = 300;
|
||||
private Semaphore concurrencySemaphore = new Semaphore(3);
|
||||
|
||||
|
||||
public GhostScriptService(OcrServiceSettings ocrServiceSettings) {
|
||||
|
||||
this.ocrServiceSettings = ocrServiceSettings;
|
||||
assertGhostscriptIsInstalled();
|
||||
}
|
||||
|
||||
|
||||
private void assertGhostscriptIsInstalled() {
|
||||
|
||||
try {
|
||||
Process p = Runtime.getRuntime().exec("gs -v");
|
||||
InputStream stdOut = p.getInputStream();
|
||||
InputStream errOut = p.getErrorStream();
|
||||
assert p.waitFor(1, TimeUnit.SECONDS);
|
||||
log.info("Ghostscript is installed.");
|
||||
String out = new String(stdOut.readAllBytes());
|
||||
String error = new String(errOut.readAllBytes());
|
||||
for (String line : out.split("\n")) {
|
||||
log.info(line);
|
||||
}
|
||||
if (!error.isBlank()) {
|
||||
log.error(error);
|
||||
}
|
||||
} catch (Exception e) {
|
||||
log.error("Ghostscript is not installed!");
|
||||
log.error(e.getMessage(), e);
|
||||
throw new RuntimeException(e);
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@SneakyThrows
|
||||
public void startBatchRender(PageBatch batch, ImageProcessingSupervisor supervisor, Consumer<ImageFile> successHandler, Consumer<String> errorHandler) {
|
||||
public void startBatchRender(PageBatch batch, ImageProcessingSupervisor supervisor, Path renderedImageDir, Consumer<ImageFile> successHandler, Consumer<String> errorHandler) {
|
||||
|
||||
supervisor.requireNoErrors();
|
||||
|
||||
List<ImageFile> renderedImageFiles = batch.getRenderedImageFiles();
|
||||
if (ocrServiceSettings.isUseCaches() && renderedImageFiles.stream()
|
||||
.allMatch(ImageFile::exists)) {
|
||||
log.info("Batch {}: Using cached GhostScript rendering with page(s) {}", batch.getIndex(), batch);
|
||||
renderedImageFiles.forEach(successHandler);
|
||||
return;
|
||||
}
|
||||
|
||||
concurrencySemaphore.acquire();
|
||||
log.info("Batch {}: starting GhostScript rendering with page(s) {}", batch.getIndex(), batch);
|
||||
executeProcess(batch, buildCmdArgs(batch, batch.getBatchDoc()), successHandler, errorHandler);
|
||||
log.info("Batch {}: starting GhostScript rendering with {} pages", batch, batch.size());
|
||||
executeProcess(batch.getIndex(), buildCmdArgs(batch, renderedImageDir, batch.getBatchDoc()), successHandler, errorHandler);
|
||||
}
|
||||
|
||||
|
||||
@SneakyThrows
|
||||
private ProcessCmdsAndRenderedImageFiles buildCmdArgs(PageBatch batch, Path document) {
|
||||
private ProcessCmdsAndRenderedImageFiles buildCmdArgs(PageBatch batch, Path outputDir, Path document) {
|
||||
|
||||
String imagePathFormat = outputDir.resolve("output_" + batch.getIndex() + ".%04d" + FORMAT).toFile().toString();
|
||||
|
||||
Map<Integer, ImageFile> fullPageImages = new HashMap<>();
|
||||
List<ImageFile> renderedImageFiles = batch.getRenderedImageFiles();
|
||||
for (int i = 1; i <= renderedImageFiles.size(); i++) {
|
||||
ImageFile renderedImageFile = renderedImageFiles.get(i - 1);
|
||||
fullPageImages.put(i, renderedImageFile);
|
||||
List<Integer> allPageNumbers = batch.getAllPageNumbers();
|
||||
|
||||
for (int i = 0; i < allPageNumbers.size(); i++) {
|
||||
Integer pageNumber = allPageNumbers.get(i);
|
||||
fullPageImages.put(i + 1, new ImageFile(pageNumber, String.format(imagePathFormat, i + 1)));
|
||||
}
|
||||
|
||||
String[] cmdArgs = buildCmdArgs(document, batch.getRenderedImageNameFormat());
|
||||
String[] cmdArgs = buildCmdArgs(document, imagePathFormat);
|
||||
|
||||
return new ProcessCmdsAndRenderedImageFiles(cmdArgs, fullPageImages);
|
||||
}
|
||||
@ -106,32 +66,16 @@ public class GhostScriptService {
|
||||
|
||||
|
||||
@SneakyThrows
|
||||
private void executeProcess(PageBatch batch, ProcessCmdsAndRenderedImageFiles processInfo, Consumer<ImageFile> successHandler, Consumer<String> errorHandler) {
|
||||
private void executeProcess(int batchIdx, ProcessCmdsAndRenderedImageFiles processInfo, Consumer<ImageFile> successHandler, Consumer<String> errorHandler) {
|
||||
|
||||
Process p = Runtime.getRuntime().exec(processInfo.cmdArgs());
|
||||
InputStream stdOut = p.getInputStream();
|
||||
GhostScriptOutputHandler stdOutLogger = GhostScriptOutputHandler.stdOut(batch.getIndex(), stdOut, processInfo.renderedPageImageFiles(), successHandler, errorHandler);
|
||||
GhostScriptOutputHandler stdOutLogger = GhostScriptOutputHandler.stdOut(batchIdx, stdOut, processInfo.renderedPageImageFiles(), successHandler, errorHandler);
|
||||
InputStream stdError = p.getErrorStream();
|
||||
GhostScriptOutputHandler stdErrorLogger = GhostScriptOutputHandler.stdError(batch.getIndex(), stdError, errorHandler);
|
||||
GhostScriptOutputHandler stdErrorLogger = GhostScriptOutputHandler.stdError(batchIdx, stdError, errorHandler);
|
||||
|
||||
stdOutLogger.start();
|
||||
stdErrorLogger.start();
|
||||
handleFinished(p, errorHandler, batch, successHandler);
|
||||
}
|
||||
|
||||
|
||||
private void handleFinished(Process p, Consumer<String> errorHandler, PageBatch batch, Consumer<ImageFile> successHandler) {
|
||||
|
||||
Thread finishedThread = new Thread(() -> {
|
||||
try {
|
||||
p.waitFor(2, TimeUnit.MINUTES);
|
||||
} catch (InterruptedException e) {
|
||||
errorHandler.accept("Batch %d: Ghostscript rendering has been terminated after 2 minutes \n %s".formatted(batch.getIndex(), e.getMessage()));
|
||||
} finally {
|
||||
concurrencySemaphore.release();
|
||||
}
|
||||
});
|
||||
finishedThread.start();
|
||||
}
|
||||
|
||||
|
||||
|
||||
@ -1,5 +1,7 @@
|
||||
package com.knecon.fforesight.service.ocr.processor.service.imageprocessing;
|
||||
|
||||
import java.nio.file.Files;
|
||||
import java.nio.file.Path;
|
||||
import java.util.List;
|
||||
import java.util.function.Consumer;
|
||||
|
||||
@ -18,6 +20,9 @@ import lombok.experimental.FieldDefaults;
|
||||
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
|
||||
public class ImageProcessingPipeline {
|
||||
|
||||
public static final String PROCESSED_DIR = "processed";
|
||||
public static final String RENDERED_DIR = "rendered";
|
||||
|
||||
GhostScriptService ghostScriptService;
|
||||
ImageProcessingService imageProcessingService;
|
||||
|
||||
@ -25,14 +30,20 @@ public class ImageProcessingPipeline {
|
||||
@SneakyThrows
|
||||
public ImageProcessingSupervisor addToPipeline(PageBatch batch) {
|
||||
|
||||
Path processedImageDir = batch.getImagePipelineDir().resolve(PROCESSED_DIR);
|
||||
Path renderedImageDir = batch.getImagePipelineDir().resolve(RENDERED_DIR);
|
||||
|
||||
Files.createDirectories(renderedImageDir);
|
||||
Files.createDirectories(processedImageDir);
|
||||
|
||||
List<Integer> pageNumbers = batch.getAllPageNumbers();
|
||||
|
||||
ImageProcessingSupervisor supervisor = new ImageProcessingSupervisor(pageNumbers);
|
||||
|
||||
Consumer<ImageFile> renderingSuccessConsumer = imageFile -> imageProcessingService.addToProcessingQueue(imageFile, batch.getProcessedImageDir(), supervisor);
|
||||
Consumer<ImageFile> renderingSuccessConsumer = imageFile -> imageProcessingService.addToProcessingQueue(imageFile, processedImageDir, supervisor);
|
||||
Consumer<String> renderingErrorConsumer = supervisor::markError;
|
||||
|
||||
ghostScriptService.startBatchRender(batch, supervisor, renderingSuccessConsumer, renderingErrorConsumer);
|
||||
ghostScriptService.startBatchRender(batch, supervisor, renderedImageDir, renderingSuccessConsumer, renderingErrorConsumer);
|
||||
|
||||
return supervisor;
|
||||
}
|
||||
|
||||
@ -1,13 +1,11 @@
|
||||
package com.knecon.fforesight.service.ocr.processor.service.imageprocessing;
|
||||
|
||||
import java.io.File;
|
||||
import java.nio.file.Path;
|
||||
import java.util.concurrent.BlockingQueue;
|
||||
import java.util.concurrent.LinkedBlockingQueue;
|
||||
|
||||
import org.springframework.stereotype.Service;
|
||||
|
||||
import com.knecon.fforesight.service.ocr.processor.OcrServiceSettings;
|
||||
import com.knecon.fforesight.service.ocr.processor.model.ImageFile;
|
||||
|
||||
import lombok.AccessLevel;
|
||||
@ -25,10 +23,9 @@ import net.sourceforge.lept4j.util.LeptUtils;
|
||||
public class ImageProcessingService {
|
||||
|
||||
BlockingQueue<ProcessParams> queue = new LinkedBlockingQueue<>();
|
||||
private final OcrServiceSettings ocrServiceSettings;
|
||||
|
||||
|
||||
public ImageProcessingService(OcrServiceSettings ocrServiceSettings) {
|
||||
public ImageProcessingService() {
|
||||
|
||||
Thread queueConsumerThread = new Thread(() -> {
|
||||
while (true) {
|
||||
@ -47,7 +44,7 @@ public class ImageProcessingService {
|
||||
}
|
||||
});
|
||||
queueConsumerThread.start();
|
||||
this.ocrServiceSettings = ocrServiceSettings;
|
||||
|
||||
}
|
||||
|
||||
|
||||
@ -61,36 +58,24 @@ public class ImageProcessingService {
|
||||
private void process(ImageFile unprocessedImage, Path outputDir, ImageProcessingSupervisor supervisor) {
|
||||
|
||||
String absoluteFilePath = outputDir.resolve(Path.of(unprocessedImage.absoluteFilePath()).getFileName()).toFile().toString();
|
||||
ImageFile processedImage = new ImageFile(unprocessedImage.pageNumber(), absoluteFilePath);
|
||||
|
||||
if (ocrServiceSettings.isUseCaches() && processedImage.exists()) {
|
||||
supervisor.markPageFinished(processedImage);
|
||||
return;
|
||||
}
|
||||
|
||||
ImageFile imageFile = new ImageFile(unprocessedImage.pageNumber(), absoluteFilePath);
|
||||
try {
|
||||
if (!unprocessedImage.exists()) {
|
||||
log.error("ERROR, rendered image {} does not exist", unprocessedImage.absoluteFilePath());
|
||||
throw new AssertionError();
|
||||
}
|
||||
synchronized (ImageProcessingSupervisor.class) {
|
||||
// Leptonica is not thread safe, but is being called in WritableOcrResultFactory as well
|
||||
Pix processedPix;
|
||||
Pix pix = unprocessedImage.readPix();
|
||||
|
||||
assert pix != null;
|
||||
|
||||
processedPix = processPix(pix);
|
||||
Leptonica1.pixWrite(processedImage.absoluteFilePath(), processedPix, ILeptonica.IFF_TIFF_PACKBITS);
|
||||
Leptonica1.pixWrite(absoluteFilePath, processedPix, ILeptonica.IFF_TIFF_PACKBITS);
|
||||
|
||||
LeptUtils.disposePix(pix);
|
||||
LeptUtils.disposePix(processedPix);
|
||||
|
||||
}
|
||||
} catch (Exception e) {
|
||||
supervisor.markError("Page %d could not be processed due to: %s".formatted(unprocessedImage.pageNumber(), e.getMessage()));
|
||||
supervisor.markError(e.getMessage());
|
||||
} finally {
|
||||
supervisor.markPageFinished(processedImage);
|
||||
log.debug("Finished page: {}", processedImage.pageNumber());
|
||||
supervisor.markPageFinished(imageFile);
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@ -69,7 +69,6 @@ public class ImageProcessingSupervisor {
|
||||
|
||||
public void markError(String errorMessage) {
|
||||
|
||||
log.error(errorMessage);
|
||||
this.errors.add(errorMessage);
|
||||
}
|
||||
|
||||
|
||||
@ -3,9 +3,13 @@ package com.knecon.fforesight.service.ocr.processor.service.imageprocessing;
|
||||
import static net.sourceforge.lept4j.ILeptonica.L_THIN_FG;
|
||||
|
||||
import java.io.Closeable;
|
||||
import java.io.IOException;
|
||||
import java.nio.IntBuffer;
|
||||
|
||||
import org.springframework.stereotype.Service;
|
||||
|
||||
import lombok.AccessLevel;
|
||||
import lombok.NoArgsConstructor;
|
||||
import lombok.experimental.FieldDefaults;
|
||||
import net.sourceforge.lept4j.Leptonica1;
|
||||
import net.sourceforge.lept4j.Pix;
|
||||
@ -15,7 +19,6 @@ import net.sourceforge.lept4j.util.LeptUtils;
|
||||
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
|
||||
public class StrokeWidthCalculator implements Closeable {
|
||||
|
||||
public static final double MINIMUM_PIXEL_DENSITY = 0.05;
|
||||
Sela thinningSel = Leptonica1.selaMakeThinSets(1, 0);
|
||||
|
||||
|
||||
@ -43,14 +46,6 @@ public class StrokeWidthCalculator implements Closeable {
|
||||
}
|
||||
|
||||
|
||||
public static boolean wordImageHasMinimumPixelDensity(Pix wordImage) {
|
||||
|
||||
IntBuffer pixelCount = IntBuffer.allocate(1);
|
||||
Leptonica1.pixCountPixels(wordImage, pixelCount, null);
|
||||
return (double) pixelCount.get(0) / (wordImage.w * wordImage.h) >= MINIMUM_PIXEL_DENSITY;
|
||||
}
|
||||
|
||||
|
||||
public boolean hasLargerStrokeWidth(Pix pix, double strokeWidth, double threshold) {
|
||||
|
||||
int roundedStrokeWidth = (int) Math.round(strokeWidth);
|
||||
|
||||
@ -1,26 +1,25 @@
|
||||
package com.knecon.fforesight.service.ocr.processor.utils;
|
||||
|
||||
import java.util.Locale;
|
||||
import org.apache.commons.lang3.StringUtils;
|
||||
|
||||
import lombok.experimental.UtilityClass;
|
||||
|
||||
@UtilityClass
|
||||
public final class OsUtils {
|
||||
|
||||
private static final String SERVICE_NAME = "azure-ocr-service";
|
||||
|
||||
|
||||
private static boolean isWindows() {
|
||||
|
||||
String osName = System.getProperty("os.name");
|
||||
if (osName == null) {
|
||||
return false;
|
||||
}
|
||||
return osName.toLowerCase(Locale.ENGLISH).contains("windows");
|
||||
return StringUtils.containsIgnoreCase(System.getProperty("os.name"), "Windows");
|
||||
}
|
||||
|
||||
|
||||
public static String getTemporaryDirectory() {
|
||||
|
||||
String tmpdir = System.getProperty("java.io.tmpdir");
|
||||
if (isWindows() && !tmpdir.isBlank()) {
|
||||
if (isWindows() && StringUtils.isNotBlank(tmpdir)) {
|
||||
return tmpdir;
|
||||
}
|
||||
return "/tmp";
|
||||
|
||||
@ -1,40 +0,0 @@
|
||||
package com.knecon.fforesight.service.ocr.processor.utils;
|
||||
|
||||
import java.util.regex.Pattern;
|
||||
|
||||
import lombok.experimental.UtilityClass;
|
||||
|
||||
@UtilityClass
|
||||
public class StringCleaningUtility {
|
||||
|
||||
public static final Pattern hyphenLineBreaks = Pattern.compile("[-~‐‒⁻−﹣゠⁓‑\\u00AD][\\r\\n]+");
|
||||
public static final Pattern linebreaks = Pattern.compile("[\\r\\n]+");
|
||||
public static final Pattern doubleWhitespaces = Pattern.compile("\\s{2,}");
|
||||
|
||||
|
||||
public static String cleanString(String value) {
|
||||
|
||||
String noHyphenLinebreaks = removeHyphenLinebreaks(value);
|
||||
String noLinebreaks = removeLinebreaks(noHyphenLinebreaks);
|
||||
return removeMultipleWhitespaces(noLinebreaks);
|
||||
}
|
||||
|
||||
|
||||
private String removeHyphenLinebreaks(String value) {
|
||||
|
||||
return hyphenLineBreaks.matcher(value).replaceAll("");
|
||||
}
|
||||
|
||||
|
||||
private String removeMultipleWhitespaces(String value) {
|
||||
|
||||
return doubleWhitespaces.matcher(value).replaceAll(" ");
|
||||
}
|
||||
|
||||
|
||||
private String removeLinebreaks(String value) {
|
||||
|
||||
return linebreaks.matcher(value).replaceAll(" ");
|
||||
}
|
||||
|
||||
}
|
||||
@ -1,14 +1,14 @@
|
||||
package com.knecon.fforesight.service.ocr.processor.visualizations;
|
||||
|
||||
import com.azure.ai.documentintelligence.models.AnalyzeResult;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.IdpResult;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.AzureAnalyzeResult;
|
||||
|
||||
import lombok.experimental.UtilityClass;
|
||||
|
||||
@UtilityClass
|
||||
public class AnalyzeResultMapper {
|
||||
|
||||
public IdpResult map(AnalyzeResult analyzeResult) {
|
||||
public AzureAnalyzeResult map(AnalyzeResult analyzeResult) {
|
||||
|
||||
return null;
|
||||
}
|
||||
|
||||
@ -1,8 +1,9 @@
|
||||
package com.knecon.fforesight.service.ocr.processor.service;
|
||||
package com.knecon.fforesight.service.ocr.processor.visualizations;
|
||||
|
||||
import java.awt.geom.AffineTransform;
|
||||
import java.awt.geom.Line2D;
|
||||
import java.awt.geom.Rectangle2D;
|
||||
import java.nio.IntBuffer;
|
||||
import java.util.ArrayList;
|
||||
import java.util.Collection;
|
||||
import java.util.Collections;
|
||||
@ -18,13 +19,13 @@ import java.util.stream.Stream;
|
||||
|
||||
import com.azure.ai.documentintelligence.models.AnalyzeResult;
|
||||
import com.azure.ai.documentintelligence.models.BoundingRegion;
|
||||
import com.azure.ai.documentintelligence.models.DocumentFontStyle;
|
||||
import com.azure.ai.documentintelligence.models.DocumentPage;
|
||||
import com.azure.ai.documentintelligence.models.DocumentSpan;
|
||||
import com.azure.ai.documentintelligence.models.DocumentStyle;
|
||||
import com.azure.ai.documentintelligence.models.DocumentTable;
|
||||
import com.azure.ai.documentintelligence.models.DocumentTableCell;
|
||||
import com.azure.ai.documentintelligence.models.DocumentWord;
|
||||
import com.azure.ai.documentintelligence.models.FontWeight;
|
||||
import com.google.common.base.Functions;
|
||||
import com.knecon.fforesight.service.ocr.processor.OcrServiceSettings;
|
||||
import com.knecon.fforesight.service.ocr.processor.model.ImageFile;
|
||||
@ -32,12 +33,10 @@ import com.knecon.fforesight.service.ocr.processor.model.PageBatch;
|
||||
import com.knecon.fforesight.service.ocr.processor.model.PageInformation;
|
||||
import com.knecon.fforesight.service.ocr.processor.model.SpanLookup;
|
||||
import com.knecon.fforesight.service.ocr.processor.model.TextPositionInImage;
|
||||
import com.knecon.fforesight.service.ocr.processor.service.imageprocessing.BBoxSnuggificationService;
|
||||
import com.knecon.fforesight.service.ocr.processor.service.imageprocessing.FontStyleDetector;
|
||||
import com.knecon.fforesight.service.ocr.processor.service.imageprocessing.ImageProcessingPipeline;
|
||||
import com.knecon.fforesight.service.ocr.processor.service.imageprocessing.ImageProcessingSupervisor;
|
||||
import com.knecon.fforesight.service.ocr.processor.service.imageprocessing.StrokeWidthCalculator;
|
||||
import com.knecon.fforesight.service.ocr.processor.visualizations.WritableOcrResult;
|
||||
import com.knecon.fforesight.service.ocr.processor.service.imageprocessing.BBoxSnuggificationService;
|
||||
import com.knecon.fforesight.service.ocr.processor.visualizations.fonts.FontMetricsProvider;
|
||||
import com.knecon.fforesight.service.ocr.processor.visualizations.fonts.FontStyle;
|
||||
import com.knecon.fforesight.service.ocr.processor.visualizations.fonts.Type0FontMetricsProvider;
|
||||
@ -57,7 +56,7 @@ import net.sourceforge.lept4j.util.LeptUtils;
|
||||
|
||||
@Slf4j
|
||||
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
|
||||
public class OcrResultPostProcessingPipeline {
|
||||
public class WritableOcrResultFactory {
|
||||
|
||||
@Getter
|
||||
Map<Integer, AffineTransform> resultToPageTransforms;
|
||||
@ -68,10 +67,10 @@ public class OcrResultPostProcessingPipeline {
|
||||
|
||||
|
||||
@SneakyThrows
|
||||
public OcrResultPostProcessingPipeline(Map<Integer, PageInformation> pageInformation,
|
||||
ImageProcessingPipeline imageProcessingPipeline,
|
||||
OcrServiceSettings settings,
|
||||
Set<AzureOcrFeature> features) {
|
||||
public WritableOcrResultFactory(Map<Integer, PageInformation> pageInformation,
|
||||
ImageProcessingPipeline imageProcessingPipeline,
|
||||
OcrServiceSettings settings,
|
||||
Set<AzureOcrFeature> features) {
|
||||
|
||||
this.imageProcessingPipeline = imageProcessingPipeline;
|
||||
this.pageInformation = pageInformation;
|
||||
@ -81,9 +80,14 @@ public class OcrResultPostProcessingPipeline {
|
||||
}
|
||||
|
||||
|
||||
public List<WritableOcrResult> processAnalyzeResult(AnalyzeResult analyzeResult, PageBatch batch) throws InterruptedException {
|
||||
public List<WritableOcrResult> buildOcrResultToWrite(AnalyzeResult analyzeResult, PageBatch batch) throws InterruptedException {
|
||||
|
||||
ImageProcessingSupervisor imageSupervisor = renderImagesIfNecessary(analyzeResult, batch);
|
||||
Map<Integer, Double> anglesPerPage = analyzeResult.getPages()
|
||||
.stream()
|
||||
.collect(Collectors.toMap(DocumentPage::getPageNumber, documentPage -> -documentPage.getAngle()));
|
||||
RotationCorrectionUtility.rotatePages(batch.getBatchDoc(), batch.getBatchDoc(), anglesPerPage);
|
||||
|
||||
ImageProcessingSupervisor imageSupervisor = imageProcessingPipeline.addToPipeline(batch);
|
||||
|
||||
List<WritableOcrResult> writableOcrResultList = new ArrayList<>();
|
||||
|
||||
@ -106,35 +110,10 @@ public class OcrResultPostProcessingPipeline {
|
||||
writableOcrResultList.add(builder.build());
|
||||
|
||||
}
|
||||
log.debug("Batch {}: finished post-processing.", batch.getIndex());
|
||||
return writableOcrResultList;
|
||||
}
|
||||
|
||||
|
||||
private ImageProcessingSupervisor renderImagesIfNecessary(AnalyzeResult analyzeResult, PageBatch batch) {
|
||||
|
||||
ImageProcessingSupervisor imageSupervisor = null;
|
||||
if (useRenderedImages()) {
|
||||
|
||||
Map<Integer, Double> anglesPerPage = analyzeResult.getPages()
|
||||
.stream()
|
||||
.collect(Collectors.toMap(DocumentPage::getPageNumber, documentPage -> -documentPage.getAngle()));
|
||||
RotationCorrectionUtility.rotatePages(batch.getBatchDoc(), batch.getBatchDoc(), anglesPerPage);
|
||||
imageSupervisor = imageProcessingPipeline.addToPipeline(batch);
|
||||
}
|
||||
return imageSupervisor;
|
||||
}
|
||||
|
||||
|
||||
private boolean useRenderedImages() {
|
||||
|
||||
if (settings.isAzureFontStyleDetection() && features.contains(AzureOcrFeature.FONT_STYLE_DETECTION)) {
|
||||
return false;
|
||||
}
|
||||
return settings.isSnuggify() || features.contains(AzureOcrFeature.FONT_STYLE_DETECTION);
|
||||
}
|
||||
|
||||
|
||||
private List<TextPositionInImage> buildTextPositionsInImage(PageBatch pageOffset,
|
||||
DocumentPage resultPage,
|
||||
AffineTransform resultToPageTransform,
|
||||
@ -142,7 +121,7 @@ public class OcrResultPostProcessingPipeline {
|
||||
PageInformation pageInformation,
|
||||
ImageProcessingSupervisor imageSupervisor) throws InterruptedException {
|
||||
|
||||
if (!useRenderedImages()) {
|
||||
if (!settings.isSnuggify() && !features.contains(AzureOcrFeature.FONT_STYLE_DETECTION)) {
|
||||
return buildText(resultPage, resultToPageTransform, lookups, pageInformation);
|
||||
}
|
||||
|
||||
@ -206,7 +185,10 @@ public class OcrResultPostProcessingPipeline {
|
||||
continue;
|
||||
}
|
||||
|
||||
if (StrokeWidthCalculator.wordImageHasMinimumPixelDensity(wordImage)) {
|
||||
IntBuffer pixelCount = IntBuffer.allocate(1);
|
||||
Leptonica1.pixCountPixels(wordImage, pixelCount, null);
|
||||
|
||||
if (pixelCount.get(0) > 3) {
|
||||
fontStyleDetector.add(textPosition, wordImage, textPosition.getFontSizeByHeight());
|
||||
}
|
||||
|
||||
@ -323,13 +305,16 @@ public class OcrResultPostProcessingPipeline {
|
||||
return Lookups.empty();
|
||||
}
|
||||
|
||||
// Azure stopped supporting bold text detection in 1.0.0 release
|
||||
SpanLookup<DocumentSpan> boldLookup = new SpanLookup<>(Stream.empty(), Function.identity());
|
||||
SpanLookup<DocumentSpan> boldLookup = new SpanLookup<>(analyzeResult.getStyles()
|
||||
.stream()
|
||||
.filter(style -> Objects.equals(style.getFontWeight(), FontWeight.BOLD))
|
||||
.map(DocumentStyle::getSpans)
|
||||
.flatMap(Collection::stream), Function.identity());
|
||||
|
||||
SpanLookup<DocumentSpan> italicLookup = new SpanLookup<>(analyzeResult.getStyles()
|
||||
.stream()
|
||||
.filter(style -> Objects.equals(style.getFontStyle(),
|
||||
DocumentFontStyle.ITALIC))
|
||||
com.azure.ai.documentintelligence.models.FontStyle.ITALIC))
|
||||
.map(DocumentStyle::getSpans)
|
||||
.flatMap(Collection::stream), Functions.identity());
|
||||
|
||||
@ -12,6 +12,8 @@ import com.azure.ai.documentintelligence.models.DocumentBarcode;
|
||||
import com.azure.ai.documentintelligence.models.DocumentFigure;
|
||||
import com.azure.ai.documentintelligence.models.DocumentKeyValuePair;
|
||||
import com.azure.ai.documentintelligence.models.DocumentLine;
|
||||
import com.azure.ai.documentintelligence.models.DocumentList;
|
||||
import com.azure.ai.documentintelligence.models.DocumentListItem;
|
||||
import com.azure.ai.documentintelligence.models.DocumentParagraph;
|
||||
import com.azure.ai.documentintelligence.models.DocumentSection;
|
||||
import com.azure.ai.documentintelligence.models.DocumentTable;
|
||||
@ -21,8 +23,8 @@ import com.azure.ai.documentintelligence.models.DocumentWord;
|
||||
import com.azure.ai.documentintelligence.models.ParagraphRole;
|
||||
import com.knecon.fforesight.service.ocr.processor.model.PageBatch;
|
||||
import com.knecon.fforesight.service.ocr.processor.model.SpanLookup;
|
||||
import com.knecon.fforesight.service.ocr.processor.visualizations.utils.LineUtils;
|
||||
import com.knecon.fforesight.service.ocr.processor.visualizations.utils.Rectangle2DBBoxCollector;
|
||||
import com.knecon.fforesight.service.ocr.processor.visualizations.utils.LineUtils;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.QuadPoint;
|
||||
import com.knecon.fforesight.service.viewerdoc.layers.IdpLayerConfig;
|
||||
import com.knecon.fforesight.service.viewerdoc.model.ColoredLine;
|
||||
@ -67,6 +69,14 @@ public class IdpLayer extends IdpLayerConfig {
|
||||
}
|
||||
|
||||
|
||||
public void addList(DocumentList list, PageBatch pageOffset) {
|
||||
|
||||
for (DocumentListItem item : list.getItems()) {
|
||||
addBoundingRegion(item.getBoundingRegions(), lists, PARAGRAPH_COLOR, pageOffset);
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
public void addBarcode(int pageNumber, DocumentBarcode barcode) {
|
||||
|
||||
addPolygon(pageNumber, barcode.getPolygon(), barcodes, IMAGE_COLOR);
|
||||
@ -75,11 +85,8 @@ public class IdpLayer extends IdpLayerConfig {
|
||||
|
||||
public void addKeyValue(DocumentKeyValuePair keyValue, PageBatch pageOffset) {
|
||||
|
||||
if (keyValue.getKey() == null || keyValue.getKey().getContent().isEmpty()) {
|
||||
return;
|
||||
}
|
||||
addBoundingRegion(keyValue.getKey().getBoundingRegions(), keyValuePairs, KEY_COLOR, pageOffset);
|
||||
if (keyValue.getValue() != null && !keyValue.getValue().getContent().isEmpty()) {
|
||||
if (keyValue.getValue() != null) {
|
||||
addBoundingRegion(keyValue.getValue().getBoundingRegions(), keyValuePairs, VALUE_COLOR, pageOffset);
|
||||
|
||||
if (keyValue.getKey().getBoundingRegions()
|
||||
|
||||
@ -65,6 +65,10 @@ public class IdpLayerFactory {
|
||||
analyzeResult.getTables()
|
||||
.forEach(documentTable -> idpLayer.addTable(documentTable, pageOffset));
|
||||
}
|
||||
if (analyzeResult.getLists() != null) {
|
||||
analyzeResult.getLists()
|
||||
.forEach(list -> idpLayer.addList(list, pageOffset));
|
||||
}
|
||||
if (analyzeResult.getKeyValuePairs() != null) {
|
||||
analyzeResult.getKeyValuePairs()
|
||||
.forEach(keyValue -> idpLayer.addKeyValue(keyValue, pageOffset));
|
||||
|
||||
@ -1,241 +0,0 @@
|
||||
package com.knecon.fforesight.service.ocr.processor.visualizations.layers;
|
||||
|
||||
import static com.knecon.fforesight.service.ocr.processor.utils.StringCleaningUtility.cleanString;
|
||||
|
||||
import java.awt.geom.AffineTransform;
|
||||
import java.util.LinkedList;
|
||||
import java.util.List;
|
||||
import java.util.Map;
|
||||
import java.util.Objects;
|
||||
import java.util.Set;
|
||||
|
||||
import com.azure.ai.documentintelligence.models.AnalyzeResult;
|
||||
import com.azure.ai.documentintelligence.models.BoundingRegion;
|
||||
import com.azure.ai.documentintelligence.models.DocumentCaption;
|
||||
import com.azure.ai.documentintelligence.models.DocumentFigure;
|
||||
import com.azure.ai.documentintelligence.models.DocumentFootnote;
|
||||
import com.azure.ai.documentintelligence.models.DocumentKeyValuePair;
|
||||
import com.azure.ai.documentintelligence.models.DocumentTable;
|
||||
import com.azure.ai.documentintelligence.models.DocumentTableCell;
|
||||
import com.knecon.fforesight.service.ocr.processor.model.DocumentSpanLookup;
|
||||
import com.knecon.fforesight.service.ocr.processor.model.PageBatch;
|
||||
import com.knecon.fforesight.service.ocr.processor.model.PageInformation;
|
||||
import com.knecon.fforesight.service.ocr.processor.visualizations.utils.Rectangle2DBBoxCollector;
|
||||
import com.knecon.fforesight.service.ocr.processor.visualizations.utils.RotationCorrectionUtility;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.AzureOcrFeature;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.Figure;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.IdpResult;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.KeyValuePair;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.QuadPoint;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.Region;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.Table;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.TableCell;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.TableCellType;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.TextRegion;
|
||||
|
||||
import lombok.AccessLevel;
|
||||
import lombok.Getter;
|
||||
import lombok.experimental.FieldDefaults;
|
||||
|
||||
@Getter
|
||||
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
|
||||
public class IdpResultFactory {
|
||||
|
||||
IdpResult idpResult;
|
||||
Map<Integer, AffineTransform> resultToPageTransforms;
|
||||
Map<Integer, PageInformation> pageInformation;
|
||||
Map<Integer, Double> angles;
|
||||
boolean rotationCorrection;
|
||||
|
||||
|
||||
public IdpResultFactory(Map<Integer, AffineTransform> resultToPageTransforms,
|
||||
Map<Integer, PageInformation> pageInformation,
|
||||
Map<Integer, Double> angles,
|
||||
Set<AzureOcrFeature> features) {
|
||||
|
||||
this.angles = angles;
|
||||
|
||||
this.rotationCorrection = features.contains(AzureOcrFeature.ROTATION_CORRECTION);
|
||||
this.resultToPageTransforms = resultToPageTransforms;
|
||||
this.pageInformation = pageInformation;
|
||||
this.idpResult = IdpResult.initSynchronized();
|
||||
}
|
||||
|
||||
|
||||
public AffineTransform getResultToPageTransform(Integer pageNumber) {
|
||||
|
||||
AffineTransform transform;
|
||||
if (rotationCorrection) {
|
||||
PageInformation page = pageInformation.get(pageNumber);
|
||||
transform = RotationCorrectionUtility.buildTransform(-angles.get(pageNumber), page.cropBox().getWidth(), page.cropBox().getHeight(), false);
|
||||
} else {
|
||||
transform = new AffineTransform();
|
||||
}
|
||||
transform.concatenate(resultToPageTransforms.get(pageNumber));
|
||||
return transform;
|
||||
}
|
||||
|
||||
|
||||
public void addAnalyzeResult(AnalyzeResult analyzeResult, PageBatch batch) {
|
||||
|
||||
DocumentSpanLookup words = new DocumentSpanLookup(analyzeResult);
|
||||
if (analyzeResult.getTables() != null) {
|
||||
analyzeResult.getTables()
|
||||
.forEach(documentTable -> addTable(documentTable, words, batch));
|
||||
}
|
||||
if (analyzeResult.getKeyValuePairs() != null) {
|
||||
analyzeResult.getKeyValuePairs()
|
||||
.forEach(documentKeyValuePair -> addKeyValuePair(documentKeyValuePair, batch));
|
||||
}
|
||||
if (analyzeResult.getFigures() != null) {
|
||||
analyzeResult.getFigures()
|
||||
.forEach(documentFigure -> addFigure(documentFigure, batch, words));
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
private void addFigure(DocumentFigure documentFigure, PageBatch batch, DocumentSpanLookup words) {
|
||||
|
||||
List<TextRegion> footNotes = new LinkedList<>();
|
||||
if (documentFigure.getFootnotes() != null) {
|
||||
documentFigure.getFootnotes()
|
||||
.stream()
|
||||
.map(footNote -> toTextRegion(footNote, batch))
|
||||
.filter(Objects::nonNull)
|
||||
.forEach(footNotes::add);
|
||||
}
|
||||
int batchPageNumber = documentFigure.getBoundingRegions()
|
||||
.get(0).getPageNumber();
|
||||
Region bbox = toRegionFromRegions(batch.getPageNumber(batchPageNumber), documentFigure.getBoundingRegions());
|
||||
TextRegion caption = toTextRegion(documentFigure.getCaption(), batch);
|
||||
idpResult.figures().add(new Figure(caption, bbox, footNotes));
|
||||
|
||||
}
|
||||
|
||||
|
||||
private void addKeyValuePair(DocumentKeyValuePair documentKeyValuePair, PageBatch batch) {
|
||||
|
||||
TextRegion key = null;
|
||||
if (documentKeyValuePair.getKey() != null && !documentKeyValuePair.getKey().getContent().isEmpty()) {
|
||||
Region region = toRegionFromRegions(batch, documentKeyValuePair.getKey().getBoundingRegions());
|
||||
key = new TextRegion(region, cleanString(documentKeyValuePair.getKey().getContent()));
|
||||
}
|
||||
TextRegion value = null;
|
||||
if (documentKeyValuePair.getValue() != null && !documentKeyValuePair.getValue().getContent().isEmpty()) {
|
||||
Region region = toRegionFromRegions(batch, documentKeyValuePair.getValue().getBoundingRegions());
|
||||
value = new TextRegion(region, cleanString(documentKeyValuePair.getValue().getContent()));
|
||||
}
|
||||
|
||||
idpResult.keyValuePairs().add(new KeyValuePair(key, value));
|
||||
}
|
||||
|
||||
|
||||
private void addTable(DocumentTable documentTable, DocumentSpanLookup words, PageBatch batch) {
|
||||
|
||||
TextRegion caption = toTextRegion(documentTable.getCaption(), batch);
|
||||
List<TableCell> tableCells = documentTable.getCells()
|
||||
.stream()
|
||||
.map(documentTableCell -> toTableCell(documentTableCell, words, batch))
|
||||
.toList();
|
||||
List<TextRegion> footNotes = new LinkedList<>();
|
||||
|
||||
if (documentTable.getFootnotes() != null) {
|
||||
documentTable.getFootnotes()
|
||||
.stream()
|
||||
.map(footNote -> toTextRegion(footNote, batch))
|
||||
.filter(Objects::nonNull)
|
||||
.forEach(footNotes::add);
|
||||
}
|
||||
List<Region> bbox = documentTable.getBoundingRegions()
|
||||
.stream()
|
||||
.map(b -> toRegion(b, batch))
|
||||
.toList();
|
||||
Table table = new Table(caption, documentTable.getColumnCount(), documentTable.getRowCount(), tableCells, footNotes, bbox);
|
||||
idpResult.tables().add(table);
|
||||
}
|
||||
|
||||
|
||||
private TextRegion toTextRegion(DocumentFootnote footNote, PageBatch batch) {
|
||||
|
||||
if (footNote == null || footNote.getBoundingRegions().isEmpty()) {
|
||||
return null;
|
||||
}
|
||||
|
||||
Region region = toRegionFromRegions(batch, footNote.getBoundingRegions());
|
||||
return new TextRegion(region, cleanString(footNote.getContent()));
|
||||
}
|
||||
|
||||
|
||||
private TextRegion toTextRegion(DocumentCaption caption, PageBatch batch) {
|
||||
|
||||
if (caption == null || caption.getBoundingRegions().isEmpty()) {
|
||||
return null;
|
||||
}
|
||||
|
||||
Region region = toRegionFromRegions(batch, caption.getBoundingRegions());
|
||||
return new TextRegion(region, cleanString(caption.getContent()));
|
||||
}
|
||||
|
||||
|
||||
private TableCell toTableCell(DocumentTableCell documentTableCell, DocumentSpanLookup words, PageBatch batch) {
|
||||
|
||||
int batchPageNumber = documentTableCell.getBoundingRegions()
|
||||
.get(0).getPageNumber();
|
||||
Region region = toRegionFromRegions(batch.getPageNumber(batchPageNumber), documentTableCell.getBoundingRegions());
|
||||
TableCellType kind = mapTableCellType(documentTableCell);
|
||||
return new TableCell(new TextRegion(region, cleanString(documentTableCell.getContent())), documentTableCell.getRowIndex(), documentTableCell.getColumnIndex(), kind);
|
||||
}
|
||||
|
||||
|
||||
private static TableCellType mapTableCellType(DocumentTableCell documentTableCell) {
|
||||
|
||||
if (documentTableCell.getKind() == null) {
|
||||
return TableCellType.CONTENT;
|
||||
}
|
||||
return switch (documentTableCell.getKind().toString()) {
|
||||
case "columnHeader" -> TableCellType.COLUMN_HEADER;
|
||||
case "rowHeader" -> TableCellType.ROW_HEADER;
|
||||
case "description" -> TableCellType.DESCRIPTION;
|
||||
case "stubHead" -> TableCellType.STUB_HEAD;
|
||||
default -> TableCellType.CONTENT;
|
||||
};
|
||||
}
|
||||
|
||||
|
||||
private Region toRegion(BoundingRegion boundingRegion, PageBatch batch) {
|
||||
|
||||
int pageNumber = batch.getPageNumber(boundingRegion.getPageNumber());
|
||||
QuadPoint qp = QuadPoint.fromPolygons(boundingRegion.getPolygon()).getTransformed(getResultToPageTransform(pageNumber));
|
||||
return new Region(pageNumber, qp.toData());
|
||||
}
|
||||
|
||||
|
||||
private Region toRegionFromRegions(int pageNumber, List<BoundingRegion> regions) {
|
||||
|
||||
if (regions.size() == 1) {
|
||||
return new Region(pageNumber, QuadPoint.fromPolygons(regions.get(0).getPolygon()).getTransformed(getResultToPageTransform(pageNumber)).toData());
|
||||
}
|
||||
QuadPoint bbox = QuadPoint.fromRectangle2D(regions.stream()
|
||||
.map(BoundingRegion::getPolygon)
|
||||
.map(QuadPoint::fromPolygons)
|
||||
.map(qp -> qp.getTransformed(getResultToPageTransform(pageNumber)).getBounds2D())
|
||||
.collect(new Rectangle2DBBoxCollector()));
|
||||
|
||||
return new Region(pageNumber, bbox.toData());
|
||||
}
|
||||
|
||||
|
||||
private Region toRegionFromRegions(PageBatch batch, List<BoundingRegion> regions) {
|
||||
|
||||
assert !regions.isEmpty();
|
||||
int batchPageNumber = regions.get(0).getPageNumber();
|
||||
if (!regions.stream()
|
||||
.map(BoundingRegion::getPageNumber)
|
||||
.allMatch(number -> number == batchPageNumber)) {
|
||||
throw new AssertionError();
|
||||
}
|
||||
int pageNumber = batch.getPageNumber(batchPageNumber);
|
||||
return toRegionFromRegions(pageNumber, regions);
|
||||
}
|
||||
|
||||
}
|
||||
@ -8,31 +8,27 @@ import java.util.Map;
|
||||
import java.util.Set;
|
||||
|
||||
import com.azure.ai.documentintelligence.models.AnalyzeResult;
|
||||
import com.knecon.fforesight.service.ocr.processor.OcrServiceSettings;
|
||||
import com.knecon.fforesight.service.ocr.processor.model.PageBatch;
|
||||
import com.knecon.fforesight.service.ocr.processor.model.PageInformation;
|
||||
import com.knecon.fforesight.service.ocr.processor.service.OcrExecutionSupervisor;
|
||||
import com.knecon.fforesight.service.ocr.processor.service.OcrResultPostProcessingPipeline;
|
||||
import com.knecon.fforesight.service.ocr.processor.OcrServiceSettings;
|
||||
import com.knecon.fforesight.service.ocr.processor.service.imageprocessing.ImageProcessingPipeline;
|
||||
import com.knecon.fforesight.service.ocr.processor.visualizations.WritableOcrResult;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.IdpResult;
|
||||
import com.knecon.fforesight.service.ocr.processor.visualizations.WritableOcrResultFactory;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.AzureOcrFeature;
|
||||
import com.knecon.fforesight.service.viewerdoc.layers.LayerGroup;
|
||||
|
||||
import lombok.AccessLevel;
|
||||
import lombok.experimental.FieldDefaults;
|
||||
import lombok.extern.slf4j.Slf4j;
|
||||
|
||||
@Slf4j
|
||||
@FieldDefaults(makeFinal = true, level = AccessLevel.PRIVATE)
|
||||
public class LayerFactory {
|
||||
|
||||
OcrExecutionSupervisor supervisor;
|
||||
OcrResultPostProcessingPipeline ocrResultPostProcessingPipeline;
|
||||
WritableOcrResultFactory writableOcrResultFactory;
|
||||
IdpLayerFactory idpLayerFactory;
|
||||
OcrDebugLayerFactory ocrDebugLayerFactory;
|
||||
OcrTextLayerFactory ocrTextLayerFactory;
|
||||
IdpResultFactory idpResultFactory;
|
||||
OcrServiceSettings settings;
|
||||
Set<AzureOcrFeature> features;
|
||||
Map<Integer, Double> angles;
|
||||
@ -44,21 +40,21 @@ public class LayerFactory {
|
||||
Map<Integer, PageInformation> pageInformation,
|
||||
ImageProcessingPipeline imageProcessingPipeline) {
|
||||
|
||||
this.ocrResultPostProcessingPipeline = new OcrResultPostProcessingPipeline(pageInformation, imageProcessingPipeline, settings, features);
|
||||
this.idpLayerFactory = new IdpLayerFactory(ocrResultPostProcessingPipeline.getResultToPageTransforms());
|
||||
this.writableOcrResultFactory = new WritableOcrResultFactory(pageInformation, imageProcessingPipeline, settings, features);
|
||||
this.idpLayerFactory = new IdpLayerFactory(writableOcrResultFactory.getResultToPageTransforms());
|
||||
this.ocrDebugLayerFactory = new OcrDebugLayerFactory();
|
||||
this.ocrTextLayerFactory = new OcrTextLayerFactory();
|
||||
this.settings = settings;
|
||||
this.features = features;
|
||||
this.supervisor = supervisor;
|
||||
this.angles = Collections.synchronizedMap(new HashMap<>());
|
||||
this.idpResultFactory = new IdpResultFactory(ocrResultPostProcessingPipeline.getResultToPageTransforms(), pageInformation, angles, features);
|
||||
}
|
||||
|
||||
|
||||
public void processAnalyzeResult(PageBatch batch, AnalyzeResult analyzeResult) throws InterruptedException {
|
||||
|
||||
List<WritableOcrResult> results = ocrResultPostProcessingPipeline.processAnalyzeResult(analyzeResult, batch);
|
||||
List<WritableOcrResult> results = writableOcrResultFactory.buildOcrResultToWrite(analyzeResult, batch);
|
||||
|
||||
results.forEach(result -> angles.put(result.getPageNumber(), result.getAngle()));
|
||||
|
||||
ocrTextLayerFactory.addWritableOcrResult(results);
|
||||
@ -68,7 +64,6 @@ public class LayerFactory {
|
||||
}
|
||||
if (features.contains(AzureOcrFeature.IDP)) {
|
||||
idpLayerFactory.addAnalyzeResult(analyzeResult, batch);
|
||||
idpResultFactory.addAnalyzeResult(analyzeResult, batch);
|
||||
}
|
||||
|
||||
this.supervisor.finishMappingResult(batch);
|
||||
@ -87,8 +82,7 @@ public class LayerFactory {
|
||||
if (features.contains(AzureOcrFeature.IDP)) {
|
||||
debugLayers.add(idpLayerFactory.getIdpLayer());
|
||||
}
|
||||
IdpResult idpResult = features.contains(AzureOcrFeature.IDP) ? idpResultFactory.getIdpResult() : null;
|
||||
return new OcrResult(List.of(ocrTextLayer), debugLayers, angles, idpResult);
|
||||
return new OcrResult(List.of(ocrTextLayer), debugLayers, angles);
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
@ -3,9 +3,8 @@ package com.knecon.fforesight.service.ocr.processor.visualizations.layers;
|
||||
import java.util.List;
|
||||
import java.util.Map;
|
||||
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.IdpResult;
|
||||
import com.knecon.fforesight.service.viewerdoc.layers.LayerGroup;
|
||||
|
||||
public record OcrResult(List<LayerGroup> regularLayers, List<LayerGroup> debugLayers, Map<Integer, Double> anglesPerPage, IdpResult idpResult) {
|
||||
public record OcrResult(List<LayerGroup> regularLayers, List<LayerGroup> debugLayers, Map<Integer, Double> anglesPerPage) {
|
||||
|
||||
}
|
||||
@ -85,14 +85,12 @@ public class RotationCorrectionUtility {
|
||||
List<String> commands = new LinkedList<>();
|
||||
|
||||
double scale = getScalingFactor(angle, page);
|
||||
double x = page.getCropBox().getWidth() / 2;
|
||||
double y = page.getCropBox().getHeight() / 2;
|
||||
commands.add("q");
|
||||
commands.add("/%s <<>> BDC".formatted(KNECON_ROTATION_CORRECTION.markedContentName()));
|
||||
commands.add(buildMatrixCommands(AffineTransform.getTranslateInstance(x, y)));
|
||||
commands.add(buildMatrixCommands(AffineTransform.getTranslateInstance(page.getPageWidth() / 2, page.getPageHeight() / 2)));
|
||||
commands.add(buildMatrixCommands(AffineTransform.getRotateInstance(Math.toRadians(angle))));
|
||||
commands.add(buildMatrixCommands(AffineTransform.getScaleInstance(scale, scale)));
|
||||
commands.add(buildMatrixCommands(AffineTransform.getTranslateInstance(-x, -y)));
|
||||
commands.add(buildMatrixCommands(AffineTransform.getTranslateInstance(-page.getPageWidth() / 2, -page.getPageHeight() / 2)));
|
||||
commands.add("EMC");
|
||||
return String.join("\n", commands);
|
||||
}
|
||||
@ -135,33 +133,23 @@ public class RotationCorrectionUtility {
|
||||
|
||||
public static AffineTransform buildTransform(double angle, double originalWidth, double originalHeight) {
|
||||
|
||||
return buildTransform(angle, originalWidth, originalHeight, true);
|
||||
}
|
||||
|
||||
|
||||
public static AffineTransform buildTransform(double angle, double originalWidth, double originalHeight, boolean quadrantRotation) {
|
||||
|
||||
int quadrants = getQuadrantRotation(angle);
|
||||
|
||||
double h = originalHeight;
|
||||
double w = originalWidth;
|
||||
|
||||
AffineTransform quadrantRotationTransform = new AffineTransform();
|
||||
if (quadrantRotation) {
|
||||
|
||||
if (quadrants == 1 || quadrants == 3) {
|
||||
w = originalHeight;
|
||||
h = originalWidth;
|
||||
}
|
||||
|
||||
quadrantRotationTransform = switch (quadrants) {
|
||||
case 1 -> new AffineTransform(0, 1, -1, 0, h, 0);
|
||||
case 2 -> new AffineTransform(-1, 0, 0, -1, w, h);
|
||||
case 3 -> new AffineTransform(0, -1, 1, 0, w - h, h);
|
||||
default -> new AffineTransform();
|
||||
};
|
||||
if (quadrants == 1 || quadrants == 3) {
|
||||
w = originalHeight;
|
||||
h = originalWidth;
|
||||
}
|
||||
|
||||
AffineTransform quadrantRotation = switch (quadrants) {
|
||||
case 1 -> new AffineTransform(0, 1, -1, 0, h, 0);
|
||||
case 2 -> new AffineTransform(-1, 0, 0, -1, w, h);
|
||||
case 3 -> new AffineTransform(0, -1, 1, 0, w - h, h);
|
||||
default -> new AffineTransform();
|
||||
};
|
||||
|
||||
double remainder = getRemainingAngle(angle, quadrants);
|
||||
double scale = getScalingFactor(remainder, w, h);
|
||||
|
||||
@ -170,7 +158,7 @@ public class RotationCorrectionUtility {
|
||||
transform.rotate(Math.toRadians(remainder));
|
||||
transform.scale(scale, scale);
|
||||
transform.translate(-w / 2, -h / 2);
|
||||
transform.concatenate(quadrantRotationTransform);
|
||||
transform.concatenate(quadrantRotation);
|
||||
|
||||
return transform;
|
||||
}
|
||||
@ -199,7 +187,7 @@ public class RotationCorrectionUtility {
|
||||
public static double getRemainingAngle(double angle, int quadrants) {
|
||||
|
||||
double referenceAngle = 90 * quadrants;
|
||||
return (angle - referenceAngle) % 360;
|
||||
return angle - referenceAngle;
|
||||
}
|
||||
|
||||
|
||||
|
||||
@ -36,11 +36,10 @@ class ImageProcessingPipelineTest {
|
||||
@BeforeEach
|
||||
public void setup() {
|
||||
|
||||
new NativeLibrariesInitializer("demo:1650351709282:7bd235e003000000004ec28a6743e1163a085e2115de2536ab6e2cfe5a", "/home/kschuettler/software/leptonica/vcpkg/installed/x64-linux-dynamic/lib/").init();
|
||||
new NativeLibrariesInitializer("demo:1650351709282:7bd235e003000000004ec28a6743e1163a085e2115de2536ab6e2cfe5a").init();
|
||||
|
||||
OcrServiceSettings settings = new OcrServiceSettings();
|
||||
ImageProcessingService imageProcessingService = new ImageProcessingService(settings);
|
||||
GhostScriptService ghostScriptService = new GhostScriptService(settings);
|
||||
ImageProcessingService imageProcessingService = new ImageProcessingService();
|
||||
GhostScriptService ghostScriptService = new GhostScriptService();
|
||||
imageProcessingPipeline = new ImageProcessingPipeline(ghostScriptService, imageProcessingService);
|
||||
}
|
||||
|
||||
|
||||
@ -1,12 +1,17 @@
|
||||
package com.knecon.fforesight.service.ocr.processor.service;
|
||||
|
||||
import static com.knecon.fforesight.service.ocr.processor.service.OCRService.IMAGE_PIPELINE_DIR;
|
||||
import static com.knecon.fforesight.service.ocr.processor.service.imageprocessing.ImageProcessingPipeline.PROCESSED_DIR;
|
||||
|
||||
import java.awt.Color;
|
||||
import java.awt.geom.AffineTransform;
|
||||
import java.awt.geom.Line2D;
|
||||
import java.awt.geom.Rectangle2D;
|
||||
import java.io.File;
|
||||
import java.io.FileInputStream;
|
||||
import java.io.FileOutputStream;
|
||||
import java.nio.file.Path;
|
||||
import java.util.ArrayList;
|
||||
import java.util.Collections;
|
||||
import java.util.List;
|
||||
import java.util.Map;
|
||||
@ -27,6 +32,7 @@ import com.knecon.fforesight.service.ocr.processor.model.ImageFile;
|
||||
import com.knecon.fforesight.service.ocr.processor.model.PageInformation;
|
||||
import com.knecon.fforesight.service.ocr.processor.model.TextPositionInImage;
|
||||
import com.knecon.fforesight.service.ocr.processor.visualizations.WritableOcrResult;
|
||||
import com.knecon.fforesight.service.ocr.processor.visualizations.WritableOcrResultFactory;
|
||||
import com.knecon.fforesight.service.ocr.processor.visualizations.layers.OcrDebugLayerFactory;
|
||||
import com.knecon.fforesight.service.ocr.processor.visualizations.utils.RotationCorrectionUtility;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.QuadPoint;
|
||||
@ -38,8 +44,15 @@ import com.pdftron.pdf.ElementBuilder;
|
||||
import com.pdftron.pdf.ElementWriter;
|
||||
import com.pdftron.pdf.PDFDoc;
|
||||
import com.pdftron.pdf.Page;
|
||||
import com.sun.jna.Memory;
|
||||
import com.sun.jna.Native;
|
||||
import com.sun.jna.Pointer;
|
||||
import com.sun.jna.ptr.PointerByReference;
|
||||
|
||||
import lombok.SneakyThrows;
|
||||
import net.sourceforge.lept4j.Box;
|
||||
import net.sourceforge.lept4j.Boxa;
|
||||
import net.sourceforge.lept4j.util.LeptUtils;
|
||||
|
||||
@Disabled // leptonica is not available in build server
|
||||
public class SnugBoxesTest {
|
||||
@ -47,8 +60,9 @@ public class SnugBoxesTest {
|
||||
public static final int PAGE_NUMBER = 41;
|
||||
public static final Path ORIGIN_FILE = Path.of("/home/kschuettler/Dokumente/TestFiles/OCR/VV-331340-first100.pdf");
|
||||
public static final Path TEST_FOLDER = Path.of("/tmp/OCR_TEST/").resolve(ORIGIN_FILE.getFileName());
|
||||
public static final Path BATCH_FOLDER = TEST_FOLDER.resolve("batch_0");
|
||||
public static final Path DESTINATION_FILE = BATCH_FOLDER.resolve("SnugBoxesTest.pdf");
|
||||
public static final Path PROCESSED_FOLDER = TEST_FOLDER.resolve(IMAGE_PIPELINE_DIR).resolve(PROCESSED_DIR);
|
||||
public static final Path DESTINATION_FILE = TEST_FOLDER.resolve("SnugBoxesTest.pdf");
|
||||
public static final Path RESULT_FILE = TEST_FOLDER.resolve(IMAGE_PIPELINE_DIR).resolve("azure_result_0.json");
|
||||
|
||||
PDFTronViewerDocumentService viewerDocumentService = new PDFTronViewerDocumentService(null);
|
||||
|
||||
@ -56,7 +70,7 @@ public class SnugBoxesTest {
|
||||
@BeforeAll
|
||||
public static void setUp() {
|
||||
|
||||
new NativeLibrariesInitializer("demo:1650351709282:7bd235e003000000004ec28a6743e1163a085e2115de2536ab6e2cfe5a", "/home/kschuettler/software/leptonica/vcpkg/installed/x64-linux-dynamic/lib/").init();
|
||||
new NativeLibrariesInitializer("demo:1650351709282:7bd235e003000000004ec28a6743e1163a085e2115de2536ab6e2cfe5a").init();
|
||||
}
|
||||
|
||||
|
||||
@ -65,24 +79,24 @@ public class SnugBoxesTest {
|
||||
public void snugBoxes() {
|
||||
|
||||
String filePath = ORIGIN_FILE.toFile().toString();
|
||||
File file = new File(filePath);
|
||||
File file = PROCESSED_FOLDER.resolve("output_0.%04d.tiff".formatted(PAGE_NUMBER)).toFile();
|
||||
assert file.exists();
|
||||
ImageFile imageFile = new ImageFile(PAGE_NUMBER, file.toString());
|
||||
AnalyzeResult result = null;
|
||||
try (var in = new FileInputStream(BATCH_FOLDER.resolve("analyzeResult.json").toFile()); JsonReader reader = DefaultJsonReader.fromStream(in, new JsonOptions());) {
|
||||
try (var in = new FileInputStream(RESULT_FILE.toFile()); JsonReader reader = DefaultJsonReader.fromStream(in, new JsonOptions());) {
|
||||
result = AnalyzeResult.fromJson(reader);
|
||||
}
|
||||
|
||||
var resultPage = result.getPages()
|
||||
.get(PAGE_NUMBER - 1);
|
||||
OcrResultPostProcessingPipeline ocrResultPostProcessingPipeline = new OcrResultPostProcessingPipeline(null, null, new OcrServiceSettings(), Set.of());
|
||||
WritableOcrResultFactory writableOcrResultFactory = new WritableOcrResultFactory(null, null, new OcrServiceSettings(), Set.of());
|
||||
OcrDebugLayerFactory debugLayerFactory = new OcrDebugLayerFactory();
|
||||
InvisibleElementRemovalService invisibleElementRemovalService = new InvisibleElementRemovalService();
|
||||
try (var in = new FileInputStream(ORIGIN_FILE.toFile()); var out = new FileOutputStream(DESTINATION_FILE.toFile())) {
|
||||
invisibleElementRemovalService.removeInvisibleElements(in, out, false);
|
||||
}
|
||||
PageInformation pageInformation = getPageInformation(PAGE_NUMBER, DESTINATION_FILE.toFile().toString());
|
||||
OcrResultPostProcessingPipeline.Lookups empty = OcrResultPostProcessingPipeline.Lookups.empty();
|
||||
WritableOcrResultFactory.Lookups empty = WritableOcrResultFactory.Lookups.empty();
|
||||
|
||||
AffineTransform pageCtm = getPageCtm(PAGE_NUMBER, filePath, resultPage.getWidth());
|
||||
// pageCtm.preConcatenate(rotationCorrection);
|
||||
@ -103,7 +117,7 @@ public class SnugBoxesTest {
|
||||
|
||||
//
|
||||
|
||||
List<TextPositionInImage> words = ocrResultPostProcessingPipeline.buildTextWithSnugBBoxes(resultPage, imageFile, pageCtm, empty, pageInformation);
|
||||
List<TextPositionInImage> words = writableOcrResultFactory.buildTextWithSnugBBoxes(resultPage, imageFile, pageCtm, empty, pageInformation);
|
||||
var results = new WritableOcrResult(PAGE_NUMBER, -resultPage.getAngle(), words, Collections.emptyList());
|
||||
debugLayerFactory.addAnalysisResult(List.of(results));
|
||||
|
||||
@ -217,7 +231,7 @@ public class SnugBoxesTest {
|
||||
@SneakyThrows
|
||||
private static AffineTransform getPageCtm(int pageNumber, String file, double imageWidh) {
|
||||
|
||||
return OcrResultPostProcessingPipeline.buildResultToPageTransform(getPageInformation(pageNumber, file), imageWidh);
|
||||
return WritableOcrResultFactory.buildResultToPageTransform(getPageInformation(pageNumber, file), imageWidh);
|
||||
}
|
||||
|
||||
|
||||
|
||||
@ -31,17 +31,12 @@ dependencies {
|
||||
implementation("org.springframework.cloud:spring-cloud-starter-openfeign:4.1.1")
|
||||
implementation("org.springframework.boot:spring-boot-starter-amqp:${springBootStarterVersion}")
|
||||
|
||||
implementation("com.iqser.red.service:persistence-service-internal-api-v1:2.224.0")
|
||||
implementation("com.knecon.fforesight:tenant-commons:0.31.0")
|
||||
implementation("com.iqser.red.commons:storage-commons:2.50.0")
|
||||
|
||||
implementation("net.logstash.logback:logstash-logback-encoder:7.4")
|
||||
implementation("ch.qos.logback:logback-classic")
|
||||
|
||||
testImplementation("org.springframework.boot:spring-boot-starter-test:${springBootStarterVersion}")
|
||||
testImplementation("com.iqser.red.commons:test-commons:2.1.0")
|
||||
testImplementation("org.springframework.amqp:spring-rabbit-test:3.0.2")
|
||||
testImplementation("com.iqser.red.commons:pdftron-logic-commons:2.32.0")
|
||||
}
|
||||
|
||||
tasks.named<BootBuildImage>("bootBuildImage") {
|
||||
|
||||
@ -9,9 +9,11 @@ import org.springframework.context.annotation.Bean;
|
||||
import org.springframework.context.annotation.Import;
|
||||
import org.springframework.scheduling.annotation.EnableAsync;
|
||||
|
||||
import com.iqser.red.storage.commons.StorageAutoConfiguration;
|
||||
import com.iqser.red.pdftronlogic.commons.InvisibleElementRemovalService;
|
||||
import com.iqser.red.pdftronlogic.commons.WatermarkRemovalService;
|
||||
import com.knecon.fforesight.service.ocr.processor.OcrServiceProcessorConfiguration;
|
||||
import com.knecon.fforesight.service.ocr.v1.server.configuration.MessagingConfiguration;
|
||||
import com.iqser.red.storage.commons.StorageAutoConfiguration;
|
||||
import com.knecon.fforesight.tenantcommons.MultiTenancyAutoConfiguration;
|
||||
|
||||
import io.micrometer.core.aop.TimedAspect;
|
||||
@ -41,4 +43,17 @@ public class Application {
|
||||
}
|
||||
|
||||
|
||||
@Bean
|
||||
public InvisibleElementRemovalService invisibleElementRemovalService() {
|
||||
|
||||
return new InvisibleElementRemovalService();
|
||||
}
|
||||
|
||||
|
||||
@Bean
|
||||
public WatermarkRemovalService watermarkRemovalService() {
|
||||
|
||||
return new WatermarkRemovalService();
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
@ -0,0 +1,11 @@
|
||||
package com.knecon.fforesight.service.ocr.v1.server.configuration;
|
||||
|
||||
import org.springframework.context.annotation.Configuration;
|
||||
|
||||
import com.knecon.fforesight.tenantcommons.queue.TenantMessagingConfiguration;
|
||||
|
||||
@Configuration
|
||||
public class TenantMessagingConfigurationImpl extends TenantMessagingConfiguration {
|
||||
|
||||
|
||||
}
|
||||
@ -1,32 +0,0 @@
|
||||
package com.knecon.fforesight.service.ocr.v1.server.configuration;
|
||||
|
||||
import static com.knecon.fforesight.service.ocr.v1.server.configuration.MessagingConfiguration.OCR_DLQ;
|
||||
import static com.knecon.fforesight.service.ocr.v1.server.configuration.MessagingConfiguration.OCR_REQUEST_EXCHANGE;
|
||||
import static com.knecon.fforesight.service.ocr.v1.server.configuration.MessagingConfiguration.OCR_REQUEST_QUEUE_PREFIX;
|
||||
|
||||
import java.util.Map;
|
||||
import java.util.Set;
|
||||
|
||||
import org.springframework.context.annotation.Bean;
|
||||
import org.springframework.context.annotation.Configuration;
|
||||
|
||||
import com.knecon.fforesight.service.ocr.v1.server.queue.OcrMessageReceiver;
|
||||
import com.knecon.fforesight.tenantcommons.model.TenantQueueConfiguration;
|
||||
import com.knecon.fforesight.tenantcommons.model.TenantQueueProvider;
|
||||
|
||||
@Configuration
|
||||
public class TenantQueueProviderConfig {
|
||||
|
||||
@Bean
|
||||
protected TenantQueueProvider getTenantQueueConfigs() {
|
||||
|
||||
return new TenantQueueProvider(Set.of(TenantQueueConfiguration.builder()
|
||||
.listenerId(OcrMessageReceiver.OCR_REQUEST_LISTENER_ID)
|
||||
.exchangeName(OCR_REQUEST_EXCHANGE)
|
||||
.queuePrefix(OCR_REQUEST_QUEUE_PREFIX)
|
||||
.dlqName(OCR_DLQ)
|
||||
.arguments(Map.of("x-max-priority", 2))
|
||||
.build()));
|
||||
}
|
||||
|
||||
}
|
||||
@ -1,13 +1,10 @@
|
||||
package com.knecon.fforesight.service.ocr.v1.server.queue;
|
||||
|
||||
import java.util.Set;
|
||||
|
||||
import org.springframework.amqp.rabbit.core.RabbitTemplate;
|
||||
import org.springframework.boot.autoconfigure.condition.ConditionalOnProperty;
|
||||
import org.springframework.stereotype.Service;
|
||||
|
||||
import com.knecon.fforesight.service.ocr.processor.service.IOcrMessageSender;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.AzureOcrFeature;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.DocumentRequest;
|
||||
import com.knecon.fforesight.service.ocr.v1.server.configuration.MessagingConfiguration;
|
||||
import com.knecon.fforesight.tenantcommons.TenantContext;
|
||||
@ -25,24 +22,24 @@ public class NoStatusUpdateOcrMessageSender implements IOcrMessageSender {
|
||||
RabbitTemplate rabbitTemplate;
|
||||
|
||||
|
||||
public void sendOcrFinished(String fileId, int totalImages, Set<AzureOcrFeature> features) {
|
||||
public void sendOcrFinished(String fileId, int totalImages) {
|
||||
|
||||
}
|
||||
|
||||
|
||||
public void sendOCRStarted(String fileId, Set<AzureOcrFeature> features) {
|
||||
public void sendOCRStarted(String fileId) {
|
||||
|
||||
}
|
||||
|
||||
|
||||
public void sendUpdate(String fileId, int finishedImages, int totalImages, Set<AzureOcrFeature> features) {
|
||||
public void sendUpdate(String fileId, int finishedImages, int totalImages) {
|
||||
|
||||
}
|
||||
|
||||
|
||||
public void sendOcrResponse(DocumentRequest request) {
|
||||
public void sendOcrResponse(String dossierId, String fileId) {
|
||||
|
||||
rabbitTemplate.convertAndSend(MessagingConfiguration.OCR_RESPONSE_EXCHANGE, TenantContext.getTenantId(), request);
|
||||
rabbitTemplate.convertAndSend(MessagingConfiguration.OCR_RESPONSE_EXCHANGE, TenantContext.getTenantId(), new DocumentRequest(dossierId, fileId));
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
@ -2,11 +2,10 @@ package com.knecon.fforesight.service.ocr.v1.server.queue;
|
||||
|
||||
import java.io.File;
|
||||
import java.io.IOException;
|
||||
import java.nio.file.Files;
|
||||
import java.nio.file.Path;
|
||||
import java.time.OffsetDateTime;
|
||||
import java.time.temporal.ChronoUnit;
|
||||
import java.util.Objects;
|
||||
import java.util.stream.Collectors;
|
||||
|
||||
import org.slf4j.MDC;
|
||||
import org.springframework.amqp.AmqpRejectAndDontRequeueException;
|
||||
@ -17,8 +16,7 @@ import org.springframework.stereotype.Service;
|
||||
import org.springframework.util.FileSystemUtils;
|
||||
|
||||
import com.fasterxml.jackson.databind.ObjectMapper;
|
||||
import com.knecon.fforesight.service.ocr.processor.utils.OsUtils;
|
||||
import com.knecon.fforesight.service.ocr.v1.server.FileStorageService;
|
||||
import com.knecon.fforesight.service.ocr.processor.service.FileStorageService;
|
||||
import com.knecon.fforesight.service.ocr.processor.service.IOcrMessageSender;
|
||||
import com.knecon.fforesight.service.ocr.processor.service.OCRService;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.DocumentRequest;
|
||||
@ -36,9 +34,6 @@ import lombok.extern.slf4j.Slf4j;
|
||||
public class OcrMessageReceiver {
|
||||
|
||||
public static final String OCR_REQUEST_LISTENER_ID = "ocr-request-listener";
|
||||
public static final String IDP_RESULT_FILE_NAME = "idpResult.json";
|
||||
public static final String VIEWER_DOCUMENT_FILE_NAME = "viewerDocument.pdf";
|
||||
public static final String DOCUMENT_FILE_NAME = "document.pdf";
|
||||
|
||||
FileStorageService fileStorageService;
|
||||
ObjectMapper objectMapper;
|
||||
@ -57,35 +52,34 @@ public class OcrMessageReceiver {
|
||||
DocumentRequest request = objectMapper.readValue(in.getBody(), DocumentRequest.class);
|
||||
String dossierId = request.getDossierId();
|
||||
String fileId = request.getFileId();
|
||||
Path runDir = Path.of(OsUtils.getTemporaryDirectory()).resolve(request.getDossierId()).resolve(request.getFileId());
|
||||
Path tmpDir = Files.createTempDirectory(null);
|
||||
|
||||
try {
|
||||
MDC.put("fileId", fileId);
|
||||
log.info("--------------------------------- Starting OCR ---------------------------------");
|
||||
log.info("Features: {}", request.getFeatures().stream().map(Objects::toString).collect(Collectors.joining(", ")));
|
||||
ocrMessageSender.sendOCRStarted(fileId, request.getFeatures());
|
||||
log.info("--------------------------------------------------------------------------");
|
||||
log.info("Starting OCR");
|
||||
|
||||
File documentFile = runDir.resolve(DOCUMENT_FILE_NAME).toFile();
|
||||
File viewerDocumentFile = runDir.resolve(VIEWER_DOCUMENT_FILE_NAME).toFile();
|
||||
File idpResultFile = runDir.resolve(IDP_RESULT_FILE_NAME).toFile();
|
||||
ocrMessageSender.sendOCRStarted(fileId);
|
||||
|
||||
File documentFile = tmpDir.resolve("document.pdf").toFile();
|
||||
File viewerDocumentFile = tmpDir.resolve("viewerDocument.pdf").toFile();
|
||||
File analyzeResultFile = tmpDir.resolve("azureAnalysisResult.json").toFile();
|
||||
|
||||
fileStorageService.downloadFiles(request, documentFile);
|
||||
|
||||
ocrService.runOcrOnDocument(dossierId, fileId, request.getFeatures(), runDir, documentFile, viewerDocumentFile, idpResultFile);
|
||||
ocrService.runOcrOnDocument(dossierId, fileId, request.getFeatures(), tmpDir, documentFile, viewerDocumentFile, analyzeResultFile);
|
||||
|
||||
fileStorageService.storeFiles(request, documentFile, viewerDocumentFile, idpResultFile);
|
||||
fileStorageService.storeFiles(request, documentFile, viewerDocumentFile, analyzeResultFile);
|
||||
|
||||
ocrMessageSender.sendOcrResponse(request);
|
||||
ocrMessageSender.sendOcrResponse(dossierId, fileId);
|
||||
} catch (Exception e) {
|
||||
log.warn("An exception occurred in ocr file stage: {}", e.getMessage());
|
||||
in.getMessageProperties().getHeaders().put(MessagingConfiguration.X_ERROR_INFO_HEADER, e.getMessage());
|
||||
in.getMessageProperties().getHeaders().put(MessagingConfiguration.X_ERROR_INFO_TIMESTAMP_HEADER, OffsetDateTime.now().truncatedTo(ChronoUnit.MILLIS));
|
||||
throw new RuntimeException(e);
|
||||
} finally {
|
||||
log.info("Done");
|
||||
log.info("--------------------------------- Done ---------------------------------");
|
||||
MDC.remove("fileId");
|
||||
FileSystemUtils.deleteRecursively(runDir);
|
||||
FileSystemUtils.deleteRecursively(tmpDir);
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@ -1,13 +1,10 @@
|
||||
package com.knecon.fforesight.service.ocr.v1.server.queue;
|
||||
|
||||
import java.util.Set;
|
||||
|
||||
import org.springframework.amqp.rabbit.core.RabbitTemplate;
|
||||
import org.springframework.boot.autoconfigure.condition.ConditionalOnProperty;
|
||||
import org.springframework.stereotype.Service;
|
||||
|
||||
import com.knecon.fforesight.service.ocr.processor.service.IOcrMessageSender;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.AzureOcrFeature;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.DocumentRequest;
|
||||
import com.knecon.fforesight.service.ocr.v1.api.model.OCRStatusUpdateResponse;
|
||||
import com.knecon.fforesight.service.ocr.v1.server.configuration.MessagingConfiguration;
|
||||
@ -28,46 +25,38 @@ public class OcrMessageSender implements IOcrMessageSender {
|
||||
RabbitTemplate rabbitTemplate;
|
||||
|
||||
|
||||
public void sendOcrFinished(String fileId, int totalImages, Set<AzureOcrFeature> features) {
|
||||
public void sendOcrFinished(String fileId, int totalImages) {
|
||||
|
||||
rabbitTemplate.convertAndSend(MessagingConfiguration.OCR_STATUS_UPDATE_RESPONSE_EXCHANGE,
|
||||
TenantContext.getTenantId(),
|
||||
OCRStatusUpdateResponse.builder()
|
||||
.fileId(fileId)
|
||||
.numberOfPagesToOCR(totalImages)
|
||||
.numberOfOCRedPages(totalImages)
|
||||
.ocrFinished(true)
|
||||
.features(features)
|
||||
.build());
|
||||
|
||||
OCRStatusUpdateResponse.builder().fileId(fileId).numberOfPagesToOCR(totalImages).numberOfOCRedPages(totalImages).ocrFinished(true).build());
|
||||
}
|
||||
|
||||
|
||||
public void sendOCRStarted(String fileId, Set<AzureOcrFeature> features) {
|
||||
public void sendOCRStarted(String fileId) {
|
||||
|
||||
rabbitTemplate.convertAndSend(MessagingConfiguration.OCR_STATUS_UPDATE_RESPONSE_EXCHANGE,
|
||||
TenantContext.getTenantId(),
|
||||
OCRStatusUpdateResponse.builder().fileId(fileId).features(features).ocrStarted(true).build());
|
||||
|
||||
OCRStatusUpdateResponse.builder().fileId(fileId).ocrStarted(true).build());
|
||||
|
||||
}
|
||||
|
||||
|
||||
public void sendUpdate(String fileId, int finishedImages, int totalImages, Set<AzureOcrFeature> features) {
|
||||
public void sendUpdate(String fileId, int finishedImages, int totalImages) {
|
||||
|
||||
rabbitTemplate.convertAndSend(MessagingConfiguration.OCR_STATUS_UPDATE_RESPONSE_EXCHANGE,
|
||||
TenantContext.getTenantId(),
|
||||
OCRStatusUpdateResponse.builder()
|
||||
.fileId(fileId)
|
||||
.features(features)
|
||||
.numberOfPagesToOCR(totalImages)
|
||||
.numberOfOCRedPages(finishedImages)
|
||||
.build());
|
||||
|
||||
OCRStatusUpdateResponse.builder().fileId(fileId).numberOfPagesToOCR(totalImages).numberOfOCRedPages(finishedImages).build());
|
||||
|
||||
}
|
||||
|
||||
|
||||
public void sendOcrResponse(DocumentRequest request) {
|
||||
public void sendOcrResponse(String dossierId, String fileId) {
|
||||
|
||||
rabbitTemplate.convertAndSend(MessagingConfiguration.OCR_RESPONSE_EXCHANGE, TenantContext.getTenantId(), request);
|
||||
rabbitTemplate.convertAndSend(MessagingConfiguration.OCR_RESPONSE_EXCHANGE, TenantContext.getTenantId(), new DocumentRequest(dossierId, fileId));
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
@ -0,0 +1,70 @@
|
||||
package com.knecon.fforesight.service.ocr.v1.server.queue;
|
||||
|
||||
import static com.knecon.fforesight.service.ocr.v1.server.configuration.MessagingConfiguration.OCR_DLQ;
|
||||
import static com.knecon.fforesight.service.ocr.v1.server.configuration.MessagingConfiguration.OCR_REQUEST_EXCHANGE;
|
||||
import static com.knecon.fforesight.service.ocr.v1.server.configuration.MessagingConfiguration.OCR_REQUEST_QUEUE_PREFIX;
|
||||
|
||||
import java.util.Map;
|
||||
import java.util.Set;
|
||||
|
||||
import org.springframework.amqp.rabbit.annotation.RabbitHandler;
|
||||
import org.springframework.amqp.rabbit.annotation.RabbitListener;
|
||||
import org.springframework.boot.context.event.ApplicationReadyEvent;
|
||||
import org.springframework.context.event.EventListener;
|
||||
import org.springframework.stereotype.Service;
|
||||
|
||||
import com.knecon.fforesight.tenantcommons.TenantProvider;
|
||||
import com.knecon.fforesight.tenantcommons.model.TenantCreatedEvent;
|
||||
import com.knecon.fforesight.tenantcommons.model.TenantQueueConfiguration;
|
||||
import com.knecon.fforesight.tenantcommons.model.TenantResponse;
|
||||
import com.knecon.fforesight.tenantcommons.queue.RabbitQueueFromExchangeService;
|
||||
import com.knecon.fforesight.tenantcommons.queue.TenantExchangeMessageReceiver;
|
||||
|
||||
@Service
|
||||
public class TenantExchangeMessageReceiverImpl extends TenantExchangeMessageReceiver {
|
||||
|
||||
public TenantExchangeMessageReceiverImpl(RabbitQueueFromExchangeService rabbitQueueService, TenantProvider tenantProvider) {
|
||||
|
||||
super(rabbitQueueService, tenantProvider);
|
||||
}
|
||||
|
||||
|
||||
@Override
|
||||
protected Set<TenantQueueConfiguration> getTenantQueueConfigs() {
|
||||
|
||||
return Set.of(TenantQueueConfiguration.builder()
|
||||
.listenerId(OcrMessageReceiver.OCR_REQUEST_LISTENER_ID)
|
||||
.exchangeName(OCR_REQUEST_EXCHANGE)
|
||||
.queuePrefix(OCR_REQUEST_QUEUE_PREFIX)
|
||||
.dlqName(OCR_DLQ)
|
||||
.arguments(Map.of("x-max-priority", 2))
|
||||
.build());
|
||||
}
|
||||
|
||||
|
||||
@EventListener(ApplicationReadyEvent.class)
|
||||
public void onApplicationReady() {
|
||||
|
||||
System.out.println("application ready invoked");
|
||||
super.initializeQueues();
|
||||
}
|
||||
|
||||
|
||||
@RabbitHandler
|
||||
@RabbitListener(queues = "#{tenantMessagingConfigurationImpl.getTenantCreatedQueueName()}")
|
||||
public void reactToTenantCreation(TenantCreatedEvent tenantCreatedEvent) {
|
||||
|
||||
super.reactToTenantCreation(tenantCreatedEvent);
|
||||
}
|
||||
|
||||
|
||||
@RabbitHandler
|
||||
@RabbitListener(queues = "#{tenantMessagingConfigurationImpl.getTenantDeletedQueueName()}")
|
||||
public void reactToTenantDeletion(TenantResponse tenantResponse) {
|
||||
|
||||
super.reactToTenantDeletion(tenantResponse);
|
||||
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
@ -1,5 +1,5 @@
|
||||
# you can list packages
|
||||
ghostscript=9.55.0~dfsg1-0ubuntu5.10
|
||||
ghostscript=9.55.0~dfsg1-0ubuntu5.9
|
||||
pkg-config
|
||||
zip
|
||||
unzip
|
||||
|
||||
@ -63,5 +63,3 @@ azure:
|
||||
|
||||
ocrService:
|
||||
sendStatusUpdates: true
|
||||
|
||||
native-libs.path: ${VCPKG_DYNAMIC_LIB}
|
||||
@ -1,50 +0,0 @@
|
||||
package com.knecon.fforesight.service.ocr.v1.api.model;
|
||||
|
||||
import static org.junit.jupiter.api.Assertions.*;
|
||||
|
||||
import java.awt.geom.Point2D;
|
||||
import java.awt.geom.Rectangle2D;
|
||||
|
||||
import org.junit.jupiter.api.Test;
|
||||
|
||||
class QuadPointTest {
|
||||
|
||||
@Test
|
||||
public void testContains() {
|
||||
|
||||
var a = new Point2D.Double(0, 0);
|
||||
var b = new Point2D.Double(0, 1);
|
||||
var c = new Point2D.Double(1, 1);
|
||||
var d = new Point2D.Double(1, 0);
|
||||
var q = new QuadPoint(a, b, c, d);
|
||||
assertTrue(q.isHorizontal());
|
||||
assertFalse(q.isVertical());
|
||||
|
||||
assertTrue(q.contains(a));
|
||||
assertTrue(q.contains(b));
|
||||
assertTrue(q.contains(c));
|
||||
assertTrue(q.contains(d));
|
||||
|
||||
var p = new Point2D.Double(0.5, 0.5);
|
||||
assertTrue(q.contains(p));
|
||||
|
||||
var r = new Rectangle2D.Double(0.5, 0.5, 0.1, 0.1);
|
||||
assertTrue(q.contains(r));
|
||||
}
|
||||
|
||||
|
||||
@Test
|
||||
public void testCenter() {
|
||||
|
||||
var a = new Point2D.Double(0, 0);
|
||||
var b = new Point2D.Double(1, 1);
|
||||
var c = new Point2D.Double(2, 1);
|
||||
var d = new Point2D.Double(1, 0);
|
||||
var q = new QuadPoint(a, b, c, d);
|
||||
assertTrue(q.isHorizontal());
|
||||
assertFalse(q.isVertical());
|
||||
assertEquals(QuadPoint.Direction.RIGHT, q.getDirection());
|
||||
assertEquals(new Point2D.Double(1, 0.5), q.getCenter());
|
||||
}
|
||||
|
||||
}
|
||||
@ -1,9 +1,6 @@
|
||||
package com.knecon.fforesight.service.ocr.v1.server;
|
||||
|
||||
import static org.assertj.core.api.Assertions.assertThat;
|
||||
import static org.mockito.ArgumentMatchers.any;
|
||||
import static org.mockito.Mockito.mock;
|
||||
import static org.mockito.Mockito.when;
|
||||
|
||||
import org.junit.jupiter.api.AfterAll;
|
||||
import org.junit.jupiter.api.AfterEach;
|
||||
@ -13,7 +10,6 @@ import org.mockito.MockitoAnnotations;
|
||||
import org.mockito.junit.jupiter.MockitoExtension;
|
||||
import org.springframework.amqp.rabbit.core.RabbitAdmin;
|
||||
import org.springframework.amqp.rabbit.core.RabbitTemplate;
|
||||
import org.springframework.amqp.rabbit.listener.MessageListenerContainer;
|
||||
import org.springframework.amqp.rabbit.listener.RabbitListenerEndpointRegistry;
|
||||
import org.springframework.beans.factory.annotation.Autowired;
|
||||
import org.springframework.boot.autoconfigure.EnableAutoConfiguration;
|
||||
@ -61,8 +57,10 @@ public class AbstractTest {
|
||||
@MockBean
|
||||
private RabbitAdmin rabbitAdmin;
|
||||
|
||||
private static String pdftronLicense;
|
||||
@MockBean
|
||||
private RabbitListenerEndpointRegistry rabbitListenerEndpointRegistry;
|
||||
|
||||
private static String pdftronLicense;
|
||||
|
||||
@BeforeEach
|
||||
public void openMocks() {
|
||||
@ -109,16 +107,6 @@ public class AbstractTest {
|
||||
@ComponentScan(excludeFilters = {@ComponentScan.Filter(type = FilterType.ASSIGNABLE_TYPE, value = StorageAutoConfiguration.class)})
|
||||
public static class TestConfiguration {
|
||||
|
||||
@Bean
|
||||
public RabbitListenerEndpointRegistry rabbitListenerEndpointRegistry() {
|
||||
|
||||
var mock = mock(RabbitListenerEndpointRegistry.class);
|
||||
when(mock.getListenerContainer(any())).thenReturn(mock(MessageListenerContainer.class));
|
||||
|
||||
return mock;
|
||||
}
|
||||
|
||||
|
||||
@Bean
|
||||
@Primary
|
||||
public StorageService inMemoryStorage() {
|
||||
|
||||
@ -1,9 +1,6 @@
|
||||
package com.knecon.fforesight.service.ocr.v1.server;
|
||||
|
||||
import static com.iqser.red.pdftronlogic.commons.PdfTextExtraction.extractAllTextFromDocument;
|
||||
import static com.knecon.fforesight.service.ocr.v1.server.queue.OcrMessageReceiver.DOCUMENT_FILE_NAME;
|
||||
import static com.knecon.fforesight.service.ocr.v1.server.queue.OcrMessageReceiver.IDP_RESULT_FILE_NAME;
|
||||
import static com.knecon.fforesight.service.ocr.v1.server.queue.OcrMessageReceiver.VIEWER_DOCUMENT_FILE_NAME;
|
||||
|
||||
import java.io.File;
|
||||
import java.io.FileInputStream;
|
||||
@ -28,12 +25,12 @@ import com.knecon.fforesight.service.ocr.v1.api.model.AzureOcrFeature;
|
||||
|
||||
import lombok.SneakyThrows;
|
||||
|
||||
// in order to run, the azure.key must be set first in the application.yml and you must set the env variable VCPKG_DYNAMIC_LIB to your tesseract and leptonica installation folder
|
||||
@Disabled
|
||||
// in order to run, the azure.key must be set first in the application.yml and you must set the env variable VCPKG_DYNAMIC_LIB to your tesseract and leptonica installation folder
|
||||
@SpringBootTest()
|
||||
public class OcrServiceIntegrationTest extends AbstractTest {
|
||||
|
||||
public static final Set<AzureOcrFeature> FEATURES = Set.of(AzureOcrFeature.ROTATION_CORRECTION, AzureOcrFeature.FONT_STYLE_DETECTION, AzureOcrFeature.IDP);
|
||||
public static final Set<AzureOcrFeature> FEATURES = Set.of(AzureOcrFeature.ROTATION_CORRECTION, AzureOcrFeature.FONT_STYLE_DETECTION);
|
||||
@Autowired
|
||||
private OCRService ocrService;
|
||||
|
||||
@ -58,7 +55,7 @@ public class OcrServiceIntegrationTest extends AbstractTest {
|
||||
@SneakyThrows
|
||||
public void testOcrWithFile() {
|
||||
|
||||
testOCR("/home/kschuettler/Dokumente/TestFiles/OCR/TestSet/VV-331340-first100.pdf");
|
||||
testOCR("/home/kschuettler/Dokumente/LayoutparsingEvaluation/RAW_FILES/Difficult Headlines/VV-284053.pdf/VV-284053.pdf.ORIGIN.pdf");
|
||||
}
|
||||
|
||||
|
||||
@ -111,9 +108,9 @@ public class OcrServiceIntegrationTest extends AbstractTest {
|
||||
|
||||
assert tmpDir.toFile().exists() || tmpDir.toFile().mkdirs();
|
||||
|
||||
var documentFile = tmpDir.resolve(Path.of(DOCUMENT_FILE_NAME));
|
||||
var viewerDocumentFile = tmpDir.resolve(Path.of(VIEWER_DOCUMENT_FILE_NAME));
|
||||
var analyzeResultFile = tmpDir.resolve(Path.of(IDP_RESULT_FILE_NAME));
|
||||
var documentFile = tmpDir.resolve(Path.of("document.pdf"));
|
||||
var viewerDocumentFile = tmpDir.resolve(Path.of("viewerDocument.pdf"));
|
||||
var analyzeResultFile = tmpDir.resolve(Path.of("azureAnalysisResult.json"));
|
||||
|
||||
Files.copy(file.toPath(), documentFile, StandardCopyOption.REPLACE_EXISTING);
|
||||
Files.copy(file.toPath(), viewerDocumentFile, StandardCopyOption.REPLACE_EXISTING);
|
||||
|
||||
@ -2,16 +2,12 @@ persistence-service.url: "http://persistence-service-v1:8080"
|
||||
|
||||
pdftron.license: demo:1650351709282:7bd235e003000000004ec28a6743e1163a085e2115de2536ab6e2cfe5a
|
||||
azure:
|
||||
endpoint: https://ff-ocr-dev.cognitiveservices.azure.com/
|
||||
key: 444fe2f83e9c48da8e588c7bd5295309 # find key in Bitwarden under: Azure IDP Test Key
|
||||
native-libs:
|
||||
|
||||
endpoint: https://ff-ocr-test.cognitiveservices.azure.com/
|
||||
key: # find key in Bitwarden under: Azure IDP Test Key
|
||||
|
||||
logging.type: ${LOGGING_TYPE:CONSOLE}
|
||||
|
||||
ocrService:
|
||||
sendStatusUpdates: false
|
||||
debug: true
|
||||
ocrService.sendStatusUpdates: false
|
||||
|
||||
management:
|
||||
endpoint:
|
||||
@ -21,5 +17,4 @@ management:
|
||||
endpoints.web.exposure.include: prometheus, health, metrics
|
||||
metrics.export.prometheus.enabled: true
|
||||
|
||||
POD_NAME: azure-ocr-service
|
||||
native-libs.path: /home/kschuettler/software/leptonica/vcpkg/installed/x64-linux-dynamic/lib/
|
||||
POD_NAME: azure-ocr-service
|
||||
@ -15,7 +15,6 @@
|
||||
<exclude name="NullAssignment"/>
|
||||
<exclude name="AssignmentInOperand"/>
|
||||
<exclude name="BeanMembersShouldSerialize"/>
|
||||
<exclude name="AvoidFieldNameMatchingMethodName"/>
|
||||
</rule>
|
||||
|
||||
</ruleset>
|
||||
|
||||
@ -17,7 +17,6 @@
|
||||
<exclude name="AssignmentInOperand"/>
|
||||
<exclude name="TestClassWithoutTestCases"/>
|
||||
<exclude name="BeanMembersShouldSerialize"/>
|
||||
<exclude name="AvoidFieldNameMatchingMethodName"/>
|
||||
</rule>
|
||||
|
||||
</ruleset>
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user