Compare commits
| Author | SHA1 | Date | |
|---|---|---|---|
 Dominique Eifländer
|
acd6d7f164 | ||
|
Dominique Eifländer
|
71025f7f16 | ||
|
Dominique Eifländer
|
ae6bad830e | ||
|
Dominique Eifländer
|
e030ec9dd2 | ||
|
Dominique Eifländer
|
49139ee603 | ||
|
Dominique Eifländer
|
07da43f2d9 | ||
|
Dominique Eifländer
|
df0bbc92c7 | ||
|
Kilian Schüttler
|
0497d764ec | ||
|
Kilian Schuettler
|
1362e4fbb2 | ||
|
Dominique Eifländer
|
665ad40b0b | ||
|
Dominique Eifländer
|
e2bcf971c9 | ||
|
Dominique Eifländer
|
dacc2f7f43 |
+3
-1
@@ -16,6 +16,8 @@ deploy:
|
||||
reports:
|
||||
dotenv: version.env
|
||||
rules:
|
||||
- if: $CI_COMMIT_BRANCH == $CI_DEFAULT_BRANCH
|
||||
- if: $CI_COMMIT_BRANCH == $CI_DEFAULT_BRANCH
|
||||
- if: $CI_COMMIT_BRANCH =~ /^release/
|
||||
- if: $CI_COMMIT_TAG
|
||||
pmd:
|
||||
allow_failure: true
|
||||
|
||||
+40
-36
@@ -83,13 +83,17 @@ public class LayoutParsingPipeline {
|
||||
|
||||
try (PDDocument originDocument = layoutParsingStorageService.getOriginFile(layoutParsingRequest.originFileStorageId())) {
|
||||
ImageServiceResponse imageServiceResponse = new ImageServiceResponse();
|
||||
if (layoutParsingRequest.imagesFileStorageId().isPresent()) {
|
||||
imageServiceResponse = layoutParsingStorageService.getImagesFile(layoutParsingRequest.imagesFileStorageId().get());
|
||||
if (layoutParsingRequest.imagesFileStorageId()
|
||||
.isPresent()) {
|
||||
imageServiceResponse = layoutParsingStorageService.getImagesFile(layoutParsingRequest.imagesFileStorageId()
|
||||
.get());
|
||||
}
|
||||
|
||||
TableServiceResponse tableServiceResponse = new TableServiceResponse();
|
||||
if (layoutParsingRequest.tablesFileStorageId().isPresent()) {
|
||||
tableServiceResponse = layoutParsingStorageService.getTablesFile(layoutParsingRequest.tablesFileStorageId().get());
|
||||
if (layoutParsingRequest.tablesFileStorageId()
|
||||
.isPresent()) {
|
||||
tableServiceResponse = layoutParsingStorageService.getTablesFile(layoutParsingRequest.tablesFileStorageId()
|
||||
.get());
|
||||
}
|
||||
|
||||
ClassificationDocument classificationDocument = parseLayout(layoutParsingRequest.layoutParsingType(), originDocument, imageServiceResponse, tableServiceResponse);
|
||||
@@ -115,25 +119,25 @@ public class LayoutParsingPipeline {
|
||||
.numberOfPages(numberOfPages)
|
||||
.duration(System.currentTimeMillis() - start)
|
||||
.message(format("""
|
||||
Layout parsing has finished in %.02f s.
|
||||
identifiers: %s
|
||||
%s
|
||||
Files have been saved with Ids:
|
||||
Structure: %s
|
||||
Text: %s
|
||||
Positions: %s
|
||||
PageData: %s
|
||||
Simplified Text: %s
|
||||
Viewer Doc: %s""",
|
||||
((float) (System.currentTimeMillis() - start)) / 1000,
|
||||
layoutParsingRequest.identifier(),
|
||||
buildSemanticNodeCountMessage(numberOfPages, documentGraph.buildSemanticNodeCounts()),
|
||||
layoutParsingRequest.structureFileStorageId(),
|
||||
layoutParsingRequest.textBlockFileStorageId(),
|
||||
layoutParsingRequest.positionBlockFileStorageId(),
|
||||
layoutParsingRequest.pageFileStorageId(),
|
||||
layoutParsingRequest.simplifiedTextStorageId(),
|
||||
layoutParsingRequest.viewerDocumentStorageId()))
|
||||
Layout parsing has finished in %.02f s.
|
||||
identifiers: %s
|
||||
%s
|
||||
Files have been saved with Ids:
|
||||
Structure: %s
|
||||
Text: %s
|
||||
Positions: %s
|
||||
PageData: %s
|
||||
Simplified Text: %s
|
||||
Viewer Doc: %s""",
|
||||
((float) (System.currentTimeMillis() - start)) / 1000,
|
||||
layoutParsingRequest.identifier(),
|
||||
buildSemanticNodeCountMessage(numberOfPages, documentGraph.buildSemanticNodeCounts()),
|
||||
layoutParsingRequest.structureFileStorageId(),
|
||||
layoutParsingRequest.textBlockFileStorageId(),
|
||||
layoutParsingRequest.positionBlockFileStorageId(),
|
||||
layoutParsingRequest.pageFileStorageId(),
|
||||
layoutParsingRequest.simplifiedTextStorageId(),
|
||||
layoutParsingRequest.viewerDocumentStorageId()))
|
||||
.build();
|
||||
}
|
||||
}
|
||||
@@ -142,14 +146,14 @@ public class LayoutParsingPipeline {
|
||||
private String buildSemanticNodeCountMessage(int numberOfPages, Map<NodeType, Long> semanticNodeCounts) {
|
||||
|
||||
return String.format("%d pages with %d sections, %d headlines, %d paragraphs, %d tables with %d cells, %d headers, and %d footers parsed",
|
||||
numberOfPages,
|
||||
semanticNodeCounts.get(NodeType.SECTION) == null ? 0 : semanticNodeCounts.get(NodeType.SECTION),
|
||||
semanticNodeCounts.get(NodeType.HEADLINE) == null ? 0 : semanticNodeCounts.get(NodeType.HEADLINE),
|
||||
semanticNodeCounts.get(NodeType.PARAGRAPH) == null ? 0 : semanticNodeCounts.get(NodeType.PARAGRAPH),
|
||||
semanticNodeCounts.get(NodeType.TABLE) == null ? 0 : semanticNodeCounts.get(NodeType.TABLE),
|
||||
semanticNodeCounts.get(NodeType.TABLE_CELL) == null ? 0 : semanticNodeCounts.get(NodeType.TABLE_CELL),
|
||||
semanticNodeCounts.get(NodeType.HEADER) == null ? 0 : semanticNodeCounts.get(NodeType.HEADER),
|
||||
semanticNodeCounts.get(NodeType.FOOTER) == null ? 0 : semanticNodeCounts.get(NodeType.FOOTER));
|
||||
numberOfPages,
|
||||
semanticNodeCounts.get(NodeType.SECTION) == null ? 0 : semanticNodeCounts.get(NodeType.SECTION),
|
||||
semanticNodeCounts.get(NodeType.HEADLINE) == null ? 0 : semanticNodeCounts.get(NodeType.HEADLINE),
|
||||
semanticNodeCounts.get(NodeType.PARAGRAPH) == null ? 0 : semanticNodeCounts.get(NodeType.PARAGRAPH),
|
||||
semanticNodeCounts.get(NodeType.TABLE) == null ? 0 : semanticNodeCounts.get(NodeType.TABLE),
|
||||
semanticNodeCounts.get(NodeType.TABLE_CELL) == null ? 0 : semanticNodeCounts.get(NodeType.TABLE_CELL),
|
||||
semanticNodeCounts.get(NodeType.HEADER) == null ? 0 : semanticNodeCounts.get(NodeType.HEADER),
|
||||
semanticNodeCounts.get(NodeType.FOOTER) == null ? 0 : semanticNodeCounts.get(NodeType.FOOTER));
|
||||
}
|
||||
|
||||
|
||||
@@ -213,7 +217,7 @@ public class LayoutParsingPipeline {
|
||||
tableExtractionService.extractTables(cleanRulings, classificationPage);
|
||||
|
||||
buildPageStatistics(classificationPage);
|
||||
increaseDocumentStatistics(classificationPage, classificationDocument);
|
||||
increaseDocumentStatistics(layoutParsingType, classificationPage, classificationDocument);
|
||||
|
||||
classificationPages.add(classificationPage);
|
||||
}
|
||||
@@ -242,11 +246,11 @@ public class LayoutParsingPipeline {
|
||||
}
|
||||
|
||||
|
||||
private void increaseDocumentStatistics(ClassificationPage classificationPage, ClassificationDocument document) {
|
||||
private void increaseDocumentStatistics(LayoutParsingType layoutParsingType, ClassificationPage classificationPage, ClassificationDocument document) {
|
||||
|
||||
if (!classificationPage.isLandscape()) {
|
||||
document.getFontSizeCounter().addAll(classificationPage.getFontSizeCounter().getCountPerValue());
|
||||
}
|
||||
if (layoutParsingType.equals(LayoutParsingType.DOCUMINE) || !classificationPage.isLandscape()) {
|
||||
document.getFontSizeCounter().addAll(classificationPage.getFontSizeCounter().getCountPerValue());
|
||||
}
|
||||
document.getFontCounter().addAll(classificationPage.getFontCounter().getCountPerValue());
|
||||
document.getTextHeightCounter().addAll(classificationPage.getTextHeightCounter().getCountPerValue());
|
||||
document.getFontStyleCounter().addAll(classificationPage.getFontStyleCounter().getCountPerValue());
|
||||
|
||||
+7
@@ -196,6 +196,12 @@ public class TextPositionSequence implements CharSequence {
|
||||
}
|
||||
|
||||
|
||||
public float getTextHeightNoPadding() {
|
||||
|
||||
return textPositions.get(0).getHeightDir();
|
||||
}
|
||||
|
||||
|
||||
@JsonIgnore
|
||||
@JsonAttribute(ignore = true)
|
||||
public float getTextHeight() {
|
||||
@@ -234,6 +240,7 @@ public class TextPositionSequence implements CharSequence {
|
||||
@JsonIgnore
|
||||
@JsonAttribute(ignore = true)
|
||||
public String getFontStyle() {
|
||||
|
||||
if (textPositions.get(0).getFontName() == null) {
|
||||
return "standard";
|
||||
}
|
||||
|
||||
+13
-10
@@ -8,6 +8,7 @@ import java.util.Map;
|
||||
|
||||
import org.springframework.stereotype.Service;
|
||||
|
||||
import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.PageInfo;
|
||||
import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableCells;
|
||||
import com.knecon.fforesight.service.layoutparser.processor.python_api.model.table.TableServiceResponse;
|
||||
|
||||
@@ -24,24 +25,26 @@ public class CvTableParsingAdapter {
|
||||
Map<Integer, List<TableCells>> tableCells = new HashMap<>();
|
||||
tableServiceResponse.getData()
|
||||
.forEach(tableData -> tableCells.computeIfAbsent(tableData.getPageInfo().getNumber(), tableCell -> new ArrayList<>())
|
||||
.addAll(convertTableCells(tableData.getTableCells())));
|
||||
.addAll(convertTableCells(tableData.getTableCells(), tableData.getPageInfo())));
|
||||
|
||||
return tableCells;
|
||||
}
|
||||
|
||||
|
||||
private Collection<TableCells> convertTableCells(List<TableCells> tableCells) {
|
||||
private Collection<TableCells> convertTableCells(List<TableCells> tableCells, PageInfo pageInfo) {
|
||||
|
||||
List<TableCells> cvParsedTableCells = new ArrayList<>();
|
||||
|
||||
tableCells.forEach(t -> cvParsedTableCells.add(TableCells.builder()
|
||||
.y0(t.getY0())
|
||||
.x1(t.getX1())
|
||||
.y1(t.getY1())
|
||||
.x0(t.getX0())
|
||||
.width(t.getWidth())
|
||||
.height(t.getHeight())
|
||||
.build()));
|
||||
tableCells.stream()
|
||||
.filter(cell -> cell.getWidth() < pageInfo.getWidth() * 0.98 && cell.getHeight() < pageInfo.getHeight() * 0.98)
|
||||
.forEach(t -> cvParsedTableCells.add(TableCells.builder()
|
||||
.y0(t.getY0())
|
||||
.x1(t.getX1())
|
||||
.y1(t.getY1())
|
||||
.x0(t.getX0())
|
||||
.width(t.getWidth())
|
||||
.height(t.getHeight())
|
||||
.build()));
|
||||
|
||||
return cvParsedTableCells;
|
||||
}
|
||||
|
||||
+2
-2
@@ -61,8 +61,8 @@ public class DocuMineBlockificationService {
|
||||
boolean newLineAfterSplit = prev != null && word.getMinYDirAdj() != prev.getMinYDirAdj() && wasSplitted && splitX1 != word.getMinXDirAdj();
|
||||
boolean isSplitByRuling = isSplitByRuling(minX, minY, maxX, maxY, word, horizontalRulingLines, verticalRulingLines);
|
||||
boolean splitByDir = prev != null && !prev.getDir().equals(word.getDir());
|
||||
boolean splitByOtherFontAndOtherY = prev != null && prev.getMaxYDirAdj() != word.getMaxYDirAdj() && (word.getFontStyle().contains("bold") && !prev.getFontStyle()
|
||||
.contains("bold") || prev.getFontStyle().contains("bold") && !word.getFontStyle().contains("bold"));
|
||||
boolean splitByOtherFontAndOtherY = prev != null && Math.abs(prev.getMaxYDirAdj() - word.getMaxYDirAdj()) > word.getTextHeight() * 0.2 && (word.getFontStyle()
|
||||
.contains("bold") && !prev.getFontStyle().contains("bold") || prev.getFontStyle().contains("bold") && !word.getFontStyle().contains("bold"));
|
||||
|
||||
Matcher matcher = pattern.matcher(chunkWords.stream().collect(Collectors.joining(" ")).toString());
|
||||
boolean startsOnSameX = Math.abs(minX - word.getMinXDirAdj()) < 5 && matcher.matches();
|
||||
|
||||
+36
-28
@@ -5,7 +5,6 @@ import java.util.Locale;
|
||||
import java.util.regex.Matcher;
|
||||
import java.util.regex.Pattern;
|
||||
|
||||
import com.knecon.fforesight.service.layoutparser.processor.utils.MarkedContentUtils;
|
||||
import org.springframework.stereotype.Service;
|
||||
|
||||
import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
|
||||
@@ -13,6 +12,7 @@ import com.knecon.fforesight.service.layoutparser.processor.model.Classification
|
||||
import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
|
||||
import com.knecon.fforesight.service.layoutparser.processor.model.PageBlockType;
|
||||
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
|
||||
import com.knecon.fforesight.service.layoutparser.processor.utils.MarkedContentUtils;
|
||||
import com.knecon.fforesight.service.layoutparser.processor.utils.PositionUtils;
|
||||
|
||||
import lombok.RequiredArgsConstructor;
|
||||
@@ -23,7 +23,7 @@ import lombok.extern.slf4j.Slf4j;
|
||||
@RequiredArgsConstructor
|
||||
public class DocuMineClassificationService {
|
||||
|
||||
private static final Pattern pattern = Pattern.compile("^(\\d{1,2}\\.){1,3}\\d{1,2}\\.?\\s[0-9A-Za-z \\[\\]]{2,50}", Pattern.CASE_INSENSITIVE);
|
||||
private static final Pattern pattern = Pattern.compile("^([1-9]\\d?\\.){1,3}\\d{1,2}\\.?\\s[0-9A-Za-z \\[\\]]{2,50}", Pattern.CASE_INSENSITIVE);
|
||||
private static final Pattern pattern2 = Pattern.compile("\\p{L}{3,}", Pattern.CASE_INSENSITIVE);
|
||||
private static final Pattern pattern3 = Pattern.compile("^(\\d{1,1}\\.){1,3}\\d{1,2}\\.?\\s[a-z]{1,2}\\/[a-z]{1,2}.*");
|
||||
|
||||
@@ -64,46 +64,54 @@ public class DocuMineClassificationService {
|
||||
return;
|
||||
}
|
||||
if (MarkedContentUtils.intersects(textBlock, page.getMarkedContentBboxPerType(), MarkedContentUtils.HEADER)
|
||||
|| PositionUtils.isOverBodyTextFrame(bodyTextFrame, textBlock, page.getRotation()) && (document.getFontSizeCounter()
|
||||
.getMostPopular() == null || textBlock.getHighestFontSize() <= document.getFontSizeCounter().getMostPopular())
|
||||
) {
|
||||
|| PositionUtils.isOverBodyTextFrame(bodyTextFrame, textBlock, page.getRotation()) && (document.getFontSizeCounter().getMostPopular() == null
|
||||
|| textBlock.getHighestFontSize() <= document.getFontSizeCounter()
|
||||
.getMostPopular())) {
|
||||
textBlock.setClassification(PageBlockType.HEADER);
|
||||
|
||||
} else if (MarkedContentUtils.intersects(textBlock, page.getMarkedContentBboxPerType(), MarkedContentUtils.FOOTER)
|
||||
|| PositionUtils.isUnderBodyTextFrame(bodyTextFrame, textBlock, page.getRotation()) && (document.getFontSizeCounter()
|
||||
.getMostPopular() == null || textBlock.getHighestFontSize() <= document.getFontSizeCounter().getMostPopular())
|
||||
) {
|
||||
|| PositionUtils.isUnderBodyTextFrame(bodyTextFrame, textBlock, page.getRotation()) && (document.getFontSizeCounter().getMostPopular() == null
|
||||
|| textBlock.getHighestFontSize() <= document.getFontSizeCounter()
|
||||
.getMostPopular())) {
|
||||
textBlock.setClassification(PageBlockType.FOOTER);
|
||||
} else if (page.getPageNumber() == 1 && (PositionUtils.getHeightDifferenceBetweenChunkWordAndDocumentWord(textBlock,
|
||||
document.getTextHeightCounter().getMostPopular()) > 2.5 && textBlock.getHighestFontSize() > document.getFontSizeCounter().getMostPopular() || page.getTextBlocks()
|
||||
.size() == 1)) {
|
||||
} else if (page.getPageNumber() == 1 && (PositionUtils.getHeightDifferenceBetweenChunkWordAndDocumentWord(textBlock, document.getTextHeightCounter().getMostPopular()) > 2.5
|
||||
&& textBlock.getHighestFontSize() > document.getFontSizeCounter().getMostPopular() || page.getTextBlocks().size() == 1)) {
|
||||
if (!Pattern.matches("[0-9]+", textBlock.toString())) {
|
||||
textBlock.setClassification(PageBlockType.TITLE);
|
||||
}
|
||||
} else if (textBlock.getText().length() > 5 && (textBlock.getMostPopularWordHeight() > document.getTextHeightCounter()
|
||||
.getMostPopular() || textBlock.getMostPopularWordFontSize() > document.getFontSizeCounter().getMostPopular()) && PositionUtils.getApproxLineCount(textBlock) < 5.9
|
||||
|
||||
&& (textBlock.getMostPopularWordStyle().contains("bold") && Character.isDigit(textBlock.toString().charAt(0)) && !matcher2.matches() && !textBlock.toString()
|
||||
.contains(":") || textBlock.toString().equals(textBlock.toString().toUpperCase(Locale.ROOT)) && !matcher2.matches() && !textBlock.toString()
|
||||
.contains(":") || textBlock.toString().startsWith("APPENDIX") || textBlock.toString().startsWith("FIGURE") || textBlock.toString()
|
||||
.startsWith("TABLE")) && !textBlock.toString().endsWith(":") && matcher2.find()) {
|
||||
} else if (textBlock.getText().length() > 5
|
||||
&& (textBlock.getMostPopularWordHeight() > document.getTextHeightCounter().getMostPopular()
|
||||
|| textBlock.getMostPopularWordFontSize() > document.getFontSizeCounter().getMostPopular())
|
||||
&& PositionUtils.getApproxLineCount(textBlock) < 5.9
|
||||
&& (textBlock.getMostPopularWordStyle().contains("bold") && Character.isDigit(textBlock.toString().charAt(0)) && matcher2.reset().find() && !textBlock.toString()
|
||||
.contains(":")
|
||||
|| textBlock.toString().equals(textBlock.toString().toUpperCase(Locale.ROOT)) && matcher2.reset().find() && !textBlock.toString().contains(":")
|
||||
|| textBlock.toString().startsWith("APPENDIX")
|
||||
|| textBlock.toString().startsWith("FIGURE")
|
||||
|| textBlock.toString().startsWith("TABLE"))
|
||||
&& !textBlock.toString().endsWith(":")
|
||||
&& matcher2.reset().find()) {
|
||||
textBlock.setClassification(PageBlockType.getHeadlineType(1));
|
||||
document.setHeadlines(true);
|
||||
|
||||
} else if (matcher.find() && PositionUtils.getApproxLineCount(textBlock) < 2.9 && matcher2.find() && !matcher3.matches()) {
|
||||
} else if (matcher.reset().find() && PositionUtils.getApproxLineCount(textBlock) < 2.9 && matcher2.reset().find() && !matcher3.reset().matches()) {
|
||||
textBlock.setClassification(PageBlockType.getHeadlineType(2));
|
||||
document.setHeadlines(true);
|
||||
} else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock) && textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter()
|
||||
.getMostPopular() && textBlock.getMostPopularWordStyle().equals("bold") && !document.getFontStyleCounter().getMostPopular().equals("bold")) {
|
||||
} else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock)
|
||||
&& textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter().getMostPopular()
|
||||
&& textBlock.getMostPopularWordStyle().equals("bold")
|
||||
&& !document.getFontStyleCounter().getMostPopular().equals("bold")) {
|
||||
textBlock.setClassification(PageBlockType.PARAGRAPH_BOLD);
|
||||
} else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock) && textBlock.getMostPopularWordFont()
|
||||
.equals(document.getFontCounter().getMostPopular()) && textBlock.getMostPopularWordStyle()
|
||||
.equals(document.getFontStyleCounter().getMostPopular()) && textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter().getMostPopular()) {
|
||||
} else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock)
|
||||
&& textBlock.getMostPopularWordFont().equals(document.getFontCounter().getMostPopular())
|
||||
&& textBlock.getMostPopularWordStyle().equals(document.getFontStyleCounter().getMostPopular())
|
||||
&& textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter().getMostPopular()) {
|
||||
textBlock.setClassification(PageBlockType.PARAGRAPH);
|
||||
} else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock) && textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter()
|
||||
.getMostPopular() && textBlock.getMostPopularWordStyle().equals("italic") && !document.getFontStyleCounter()
|
||||
.getMostPopular()
|
||||
.equals("italic") && PositionUtils.getApproxLineCount(textBlock) < 2.9) {
|
||||
} else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock)
|
||||
&& textBlock.getMostPopularWordFontSize() == document.getFontSizeCounter().getMostPopular()
|
||||
&& textBlock.getMostPopularWordStyle().equals("italic")
|
||||
&& !document.getFontStyleCounter().getMostPopular().equals("italic")
|
||||
&& PositionUtils.getApproxLineCount(textBlock) < 2.9) {
|
||||
textBlock.setClassification(PageBlockType.PARAGRAPH_ITALIC);
|
||||
} else if (PositionUtils.isWithinBodyTextFrame(bodyTextFrame, textBlock)) {
|
||||
textBlock.setClassification(PageBlockType.PARAGRAPH_UNKNOWN);
|
||||
|
||||
+11
-18
@@ -4,11 +4,9 @@ import java.awt.geom.AffineTransform;
|
||||
import java.awt.geom.Rectangle2D;
|
||||
import java.io.IOException;
|
||||
import java.io.OutputStream;
|
||||
import java.util.HashSet;
|
||||
import java.util.Set;
|
||||
|
||||
import org.apache.pdfbox.cos.COSDictionary;
|
||||
import org.apache.pdfbox.cos.COSName;
|
||||
import org.apache.pdfbox.pdfwriter.compress.CompressParameters;
|
||||
import org.apache.pdfbox.pdmodel.PDDocument;
|
||||
import org.apache.pdfbox.pdmodel.PDDocumentCatalog;
|
||||
import org.apache.pdfbox.pdmodel.PDPage;
|
||||
@@ -40,7 +38,6 @@ import lombok.extern.slf4j.Slf4j;
|
||||
@RequiredArgsConstructor
|
||||
public class ViewerDocumentService {
|
||||
|
||||
|
||||
private static final String LAYER_NAME = "Layout grid";
|
||||
private static final int FONT_SIZE = 10;
|
||||
public static final float LINE_WIDTH = 1f;
|
||||
@@ -54,8 +51,7 @@ public class ViewerDocumentService {
|
||||
LayoutGrid layoutGrid = layoutGridService.createLayoutGrid(document);
|
||||
// PDDocument.save() is very slow, since it actually traverses the entire pdf and writes a new one.
|
||||
// If we collect all COSDictionaries we changed and tell it explicitly to only add the changed ones by using saveIncremental it's very fast.
|
||||
Set<COSDictionary> dictionariesToUpdate = new HashSet<>();
|
||||
PDOptionalContentGroup layer = addLayerToDocument(pdDocument, dictionariesToUpdate, layerVisibilityDefaultValue);
|
||||
PDOptionalContentGroup layer = addLayerToDocument(pdDocument, layerVisibilityDefaultValue);
|
||||
PDFont font = new PDType1Font(Standard14Fonts.FontName.HELVETICA);
|
||||
|
||||
for (int pageNumber = 0; pageNumber < pdDocument.getNumberOfPages(); pageNumber++) {
|
||||
@@ -68,7 +64,8 @@ public class ViewerDocumentService {
|
||||
// e.g. not escaped matrix transformations.
|
||||
escapePreviousContents(pdDocument, pdPage);
|
||||
|
||||
VisualizationsOnPage visualizationsOnPage = layoutGrid.getVisualizationsPerPages().get(pageNumber);
|
||||
VisualizationsOnPage visualizationsOnPage = layoutGrid.getVisualizationsPerPages()
|
||||
.get(pageNumber);
|
||||
assert pageNumber == visualizationsOnPage.getPageNumber();
|
||||
// We need to append to the content stream, otherwise the content could be overlapped by following content.
|
||||
try (var contentStream = new PDPageContentStream(pdDocument, pdPage, PDPageContentStream.AppendMode.APPEND, true)) {
|
||||
@@ -102,11 +99,11 @@ public class ViewerDocumentService {
|
||||
contentStream.setFont(font, FONT_SIZE);
|
||||
contentStream.beginText();
|
||||
Matrix textMatrix = new Matrix((float) textDeRotationMatrix.getScaleX(),
|
||||
(float) textDeRotationMatrix.getShearX(),
|
||||
(float) textDeRotationMatrix.getShearY(),
|
||||
(float) textDeRotationMatrix.getScaleY(),
|
||||
(float) placedText.lineStart().getX(),
|
||||
(float) placedText.lineStart().getY());
|
||||
(float) textDeRotationMatrix.getShearX(),
|
||||
(float) textDeRotationMatrix.getShearY(),
|
||||
(float) textDeRotationMatrix.getScaleY(),
|
||||
(float) placedText.lineStart().getX(),
|
||||
(float) placedText.lineStart().getY());
|
||||
textMatrix.translate(-((font.getStringWidth(placedText.text()) / 1000) * FONT_SIZE + (2 * LINE_WIDTH) + 4), -FONT_SIZE);
|
||||
contentStream.setTextMatrix(textMatrix);
|
||||
contentStream.showText(placedText.text());
|
||||
@@ -115,12 +112,9 @@ public class ViewerDocumentService {
|
||||
contentStream.restoreGraphicsState();
|
||||
contentStream.endMarkedContent();
|
||||
}
|
||||
dictionariesToUpdate.add(pdPage.getCOSObject());
|
||||
dictionariesToUpdate.add(pdPage.getResources().getCOSObject());
|
||||
}
|
||||
dictionariesToUpdate.add(pdDocument.getDocumentInformation().getCOSObject());
|
||||
// dictionariesToUpdate.add(pdDocument.getDocument().getTrailer());
|
||||
pdDocument.saveIncremental(outputStream, dictionariesToUpdate);
|
||||
pdDocument.save(outputStream, CompressParameters.NO_COMPRESSION);
|
||||
}
|
||||
|
||||
|
||||
@@ -145,7 +139,7 @@ public class ViewerDocumentService {
|
||||
}
|
||||
|
||||
|
||||
private static PDOptionalContentGroup addLayerToDocument(PDDocument pdDocument, Set<COSDictionary> dictionariesToUpdate, boolean layerVisibilityDefaultValue) {
|
||||
private static PDOptionalContentGroup addLayerToDocument(PDDocument pdDocument, boolean layerVisibilityDefaultValue) {
|
||||
|
||||
PDDocumentCatalog catalog = pdDocument.getDocumentCatalog();
|
||||
PDOptionalContentProperties ocprops = catalog.getOCProperties();
|
||||
@@ -161,7 +155,6 @@ public class ViewerDocumentService {
|
||||
ocprops.addGroup(layer);
|
||||
}
|
||||
ocprops.setGroupEnabled(layer, layerVisibilityDefaultValue);
|
||||
dictionariesToUpdate.add(catalog.getCOSObject());
|
||||
return layer;
|
||||
}
|
||||
|
||||
|
||||
+10
-18
@@ -28,15 +28,13 @@ import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPosit
|
||||
*
|
||||
* @author Ben Litchfield
|
||||
*/
|
||||
public class TextPositionSequenceComparator implements Comparator<TextPositionSequence>
|
||||
{
|
||||
public class TextPositionSequenceComparator implements Comparator<TextPositionSequence> {
|
||||
|
||||
@Override
|
||||
public int compare(TextPositionSequence pos1, TextPositionSequence pos2)
|
||||
{
|
||||
public int compare(TextPositionSequence pos1, TextPositionSequence pos2) {
|
||||
// only compare text that is in the same direction
|
||||
int cmp1 = Float.compare(pos1.getDir().getDegrees(), pos2.getDir().getDegrees());
|
||||
if (cmp1 != 0)
|
||||
{
|
||||
if (cmp1 != 0) {
|
||||
return cmp1;
|
||||
}
|
||||
|
||||
@@ -48,25 +46,19 @@ public class TextPositionSequenceComparator implements Comparator<TextPositionSe
|
||||
float pos2YBottom = pos2.getMaxYDirAdj();
|
||||
|
||||
// note that the coordinates have been adjusted so 0,0 is in upper left
|
||||
float pos1YTop = pos1YBottom - pos1.getTextHeight();
|
||||
float pos2YTop = pos2YBottom - pos2.getTextHeight();
|
||||
float pos1YTop = pos1YBottom - pos1.getTextHeightNoPadding();
|
||||
float pos2YTop = pos2YBottom - pos2.getTextHeightNoPadding();
|
||||
|
||||
float yDifference = Math.abs(pos1YBottom - pos2YBottom);
|
||||
|
||||
// we will do a simple tolerance comparison
|
||||
if (yDifference < .1 ||
|
||||
pos2YBottom >= pos1YTop && pos2YBottom <= pos1YBottom ||
|
||||
pos1YBottom >= pos2YTop && pos1YBottom <= pos2YBottom)
|
||||
{
|
||||
if (yDifference < .1 || pos2YBottom >= pos1YTop && pos2YBottom <= pos1YBottom || pos1YBottom >= pos2YTop && pos1YBottom <= pos2YBottom) {
|
||||
return Float.compare(x1, x2);
|
||||
}
|
||||
else if (pos1YBottom < pos2YBottom)
|
||||
{
|
||||
} else if (pos1YBottom < pos2YBottom) {
|
||||
return -1;
|
||||
}
|
||||
else
|
||||
{
|
||||
} else {
|
||||
return 1;
|
||||
}
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
+43
-2
@@ -61,12 +61,13 @@ public class PdfSegmentationServiceTest extends AbstractTest {
|
||||
private SectionsBuilderService sectionsBuilderService;
|
||||
|
||||
|
||||
public ClassificationDocument buildClassificationDocument(PDDocument originDocument) {
|
||||
@SneakyThrows
|
||||
public ClassificationDocument buildClassificationDocument(PDDocument originDocument, TableServiceResponse tableServiceResponse) {
|
||||
|
||||
ClassificationDocument classificationDocument = layoutParsingPipeline.parseLayout(LayoutParsingType.REDACT_MANAGER,
|
||||
originDocument,
|
||||
new ImageServiceResponse(),
|
||||
new TableServiceResponse());
|
||||
tableServiceResponse);
|
||||
|
||||
redactManagerClassificationService.classifyDocument(classificationDocument);
|
||||
|
||||
@@ -76,6 +77,13 @@ public class PdfSegmentationServiceTest extends AbstractTest {
|
||||
}
|
||||
|
||||
|
||||
@SneakyThrows
|
||||
public ClassificationDocument buildClassificationDocument(PDDocument originDocument) {
|
||||
|
||||
return buildClassificationDocument(originDocument, new TableServiceResponse());
|
||||
}
|
||||
|
||||
|
||||
@Test
|
||||
public void tablesToHtmlDebugger() throws IOException {
|
||||
|
||||
@@ -88,6 +96,39 @@ public class PdfSegmentationServiceTest extends AbstractTest {
|
||||
}
|
||||
|
||||
|
||||
@Test
|
||||
public void tablesToHtmlDebuggerWithCVResponse() throws IOException {
|
||||
|
||||
ClassPathResource pdfFileResource = new ClassPathResource("files/cv_tables/ScanRotationBorder.pdf");
|
||||
ClassPathResource cvTablesResource = new ClassPathResource("files/cv_tables/ScanRotationBorder.TABLES.json");
|
||||
var tableServiceResponse = objectMapper.readValue(cvTablesResource.getInputStream(), TableServiceResponse.class);
|
||||
|
||||
ClassificationDocument document = buildClassificationDocument(Loader.loadPDF(pdfFileResource.getFile()), tableServiceResponse);
|
||||
|
||||
toHtml(document, "/tmp/ScanRotationBorder.html");
|
||||
}
|
||||
|
||||
|
||||
@Test
|
||||
public void testScanRotationBorderIsIgnored() throws IOException {
|
||||
|
||||
ClassPathResource pdfFileResource = new ClassPathResource("files/cv_tables/ScanRotationBorder.pdf");
|
||||
ClassPathResource cvTablesResource = new ClassPathResource("files/cv_tables/ScanRotationBorder.TABLES.json");
|
||||
var tableServiceResponse = objectMapper.readValue(cvTablesResource.getInputStream(), TableServiceResponse.class);
|
||||
|
||||
ClassificationDocument document = buildClassificationDocument(Loader.loadPDF(pdfFileResource.getFile()), tableServiceResponse);
|
||||
assertThat(document.getSections().stream().flatMap(paragraph -> paragraph.getTables().stream()).collect(Collectors.toList())).isNotEmpty();
|
||||
var tables = document.getSections().stream().flatMap(paragraph -> paragraph.getTables().stream()).toList();
|
||||
|
||||
// Quality of the table parsing is not good, because the file is rotated at scanning.
|
||||
// We only asset that the table border is not the page border.
|
||||
tables.forEach(table -> {
|
||||
assertThat(table.getMinX()).isGreaterThan(75);
|
||||
assertThat(table.getMaxX()).isLessThan(512);
|
||||
});
|
||||
}
|
||||
|
||||
|
||||
@Test
|
||||
@SneakyThrows
|
||||
public void testMapping() {
|
||||
|
||||
BIN
Binary file not shown.
+1
File diff suppressed because one or more lines are too long
Reference in New Issue
Block a user