akra-certificate: change column detector slightly

* introduce settings
This commit is contained in:
Kilian Schuettler 2024-05-07 15:39:35 +02:00
parent 8e7bed8b52
commit 81231ae486
4 changed files with 52 additions and 14 deletions

View File

@ -96,6 +96,7 @@ public class LayoutParsingPipeline {
VisualLayoutParsingAdapter visualLayoutParsingAdapter;
ClarifyndClassificationService clarifyndClassificationService;
GraphicExtractorService graphicExtractorService;
LayoutparserSettings settings;
public LayoutParsingFinishedEvent parseLayoutAndSaveFilesToStorage(LayoutParsingRequest layoutParsingRequest) throws IOException {
@ -104,24 +105,32 @@ public class LayoutParsingPipeline {
log.info("Starting layout parsing for {}", layoutParsingRequest.identifier());
File originFile = layoutParsingStorageService.getOriginFile(layoutParsingRequest.originFileStorageId());
File viewerDocumentFile = layoutParsingStorageService.getViewerDocFile(layoutParsingRequest.viewerDocumentStorageId()).orElse(originFile);
File viewerDocumentFile = layoutParsingStorageService.getViewerDocFile(layoutParsingRequest.viewerDocumentStorageId())
.orElse(originFile);
VisualLayoutParsingResponse visualLayoutParsingResponse = new VisualLayoutParsingResponse();
if (layoutParsingRequest.visualLayoutParsingFileId().isPresent()) {
visualLayoutParsingResponse = layoutParsingStorageService.getVisualLayoutParsingFile(layoutParsingRequest.visualLayoutParsingFileId().get());
if (layoutParsingRequest.visualLayoutParsingFileId()
.isPresent()) {
visualLayoutParsingResponse = layoutParsingStorageService.getVisualLayoutParsingFile(layoutParsingRequest.visualLayoutParsingFileId()
.get());
}
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(LayoutParsingType.CLARIFYND_PARAGRAPH_DEBUG,
ClassificationDocument classificationDocument = parseLayout(settings.getLayoutParsingTypeOverride() == null //
? layoutParsingRequest.layoutParsingType() : settings.getLayoutParsingTypeOverride(),
originFile,
imageServiceResponse,
tableServiceResponse,
@ -130,11 +139,17 @@ public class LayoutParsingPipeline {
log.info("Building document graph for {}", layoutParsingRequest.identifier());
Document documentGraph = observeBuildDocumentGraph(LayoutParsingType.CLARIFYND_PARAGRAPH_DEBUG, classificationDocument);
Document documentGraph = observeBuildDocumentGraph(settings.getLayoutParsingTypeOverride() == null //
? layoutParsingRequest.layoutParsingType() : settings.getLayoutParsingTypeOverride(), classificationDocument);
log.info("Creating viewer document for {}", layoutParsingRequest.identifier());
layoutGridService.addLayoutGrid(viewerDocumentFile, documentGraph, viewerDocumentFile, false, layoutParsingRequest.visualLayoutParsingFileId().isPresent());
layoutGridService.addLayoutGrid(viewerDocumentFile,
documentGraph,
viewerDocumentFile,
false,
layoutParsingRequest.visualLayoutParsingFileId()
.isPresent());
log.info("Storing resulting files for {}", layoutParsingRequest.identifier());
@ -224,7 +239,9 @@ public class LayoutParsingPipeline {
Map<Integer, List<ClassifiedImage>> signatures = visualLayoutParsingAdapter.buildExtractedSignaturesPerPage(visualLayoutParsingResponse);
ClassificationDocument classificationDocument = new ClassificationDocument();
classificationDocument.getVisualizations().setActive(identifier.containsKey("debug"));
if (settings.isDebug() || identifier.containsKey("debug")) {
classificationDocument.getVisualizations().setActive(true);
}
List<ClassificationPage> classificationPages = new ArrayList<>();

View File

@ -0,0 +1,20 @@
package com.knecon.fforesight.service.layoutparser.processor;
import org.springframework.boot.context.properties.ConfigurationProperties;
import org.springframework.context.annotation.Configuration;
import com.knecon.fforesight.service.layoutparser.internal.api.queue.LayoutParsingType;
import lombok.AccessLevel;
import lombok.Data;
import lombok.experimental.FieldDefaults;
@Data
@Configuration
@ConfigurationProperties("layoutparser")
@FieldDefaults(level = AccessLevel.PRIVATE)
public class LayoutparserSettings {
boolean debug;
LayoutParsingType layoutParsingTypeOverride;
}

View File

@ -16,6 +16,7 @@ public class ColumnDetector {
final static int globalEndIdx = bins_num; // i chose 7, since thirds seems a likely split for columns, therefore divided by 6 would eliminate those.
public static final double DERIVATIVE_ZERO_THRESHOLD = 1e-10;
public static final double MINIMUM_THRESHOLD_FOR_COLUMNS = 0.05;
public static final double NEAR_GLOBAL_THRESHOLD = 0.5;
double minY;
double maxY;
double midY;
@ -237,10 +238,10 @@ public class ColumnDetector {
List<Integer> nearGlobalDvMaximaIdx = new LinkedList<>();
List<Integer> nearGlobalDvMinimaIdx = new LinkedList<>();
for (int i = globalStartIdx; i < globalEndIdx; i++) {
if (derivative[i] <= minDvValue * 0.8) {
if (derivative[i] <= minDvValue * NEAR_GLOBAL_THRESHOLD) {
nearGlobalDvMinimaIdx.add(i);
}
if (derivative[i] >= maxDvValue * 0.8) {
if (derivative[i] >= maxDvValue * NEAR_GLOBAL_THRESHOLD) {
nearGlobalDvMaximaIdx.add(i);
}
}

View File

@ -68,7 +68,7 @@ public class LayoutGridService {
public void addLayoutGrid(File originFile, Document document, File destinationFile, boolean layerVisibilityDefaultValue, boolean writeVisualLayoutParsingGrid) {
List<Visualizations> allVisualizations;
Visualizations layoutGrid = this.addLayoutGrid(document, true, false);
Visualizations layoutGrid = this.addLayoutGrid(document, layerVisibilityDefaultValue, false);
if (writeVisualLayoutParsingGrid) {
Visualizations visualLayoutGrid = this.addLayoutGrid(document, layerVisibilityDefaultValue, true);
allVisualizations = Stream.concat(Stream.of(layoutGrid, visualLayoutGrid), document.getVisualizations().streamAll())