Compare commits

..
1 Commits
Author SHA1 Message Date
Timo Bejan 64a6242bae ignore mc files 2024-03-11 16:14:48 +02:00
12 changed files with 155 additions and 78 deletions
@@ -25,5 +25,4 @@ dependencies {
implementation("com.fasterxml.jackson.datatype:jackson-datatype-jsr310:${jacksonVersion}")
implementation("org.springframework.boot:spring-boot-starter-web:3.1.3")
implementation("org.jgrapht:jgrapht-core:1.5.2")
implementation("org.tinspin:tinspin-indexes:2.1.3")
}
@@ -1,10 +1,10 @@
package com.knecon.fforesight.service.layoutparser.processor.docstrum.service;
import java.util.ArrayList;
import java.util.Comparator;
import java.util.List;
import org.springframework.stereotype.Service;
import org.tinspin.index.Index;
import org.tinspin.index.kdtree.KDTree;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Character;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Neighbor;
@@ -13,23 +13,115 @@ import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Neigh
public class NearestNeighbourService {
private static final int NUMBER_OF_NEIGHBOURS = 8;
private static final double STEP = 16.0;
public void findNearestNeighbors(List<Character> characters) {
KDTree<Character> kdTree = KDTree.create(2);
characters.forEach(c -> kdTree.insert(new double[]{c.getX(), c.getY()}, c));
for(Character c : characters) {
Index.PointIteratorKnn<Character> iterator = kdTree.queryKnn(new double[]{c.getX(), c.getY()}, NUMBER_OF_NEIGHBOURS + 1);
// skip the first as this is identity
if(iterator.hasNext()) {
iterator.next();
}
while(iterator.hasNext()) {
c.getNeighbors().add(new Neighbor(iterator.next().value(), c));
}
if (characters.isEmpty() || characters.size() == 1) {
return;
}
characters.sort(Comparator.comparingDouble(Character::getX));
int maxNeighborCount = NUMBER_OF_NEIGHBOURS;
if (characters.size() <= NUMBER_OF_NEIGHBOURS) {
maxNeighborCount = characters.size() - 1;
}
for (int i = 0; i < characters.size(); i++) {
Neighbor[] candidates = new Neighbor[maxNeighborCount + 1];
int neighborInsertionIndex = 0;
int neighborCount = 0;
int start = i;
int end = i + 1;
double distance = Double.POSITIVE_INFINITY;
for (double searchDistance = 0; searchDistance < distance; ) {
searchDistance += STEP;
boolean newCandidatesFound = false;
while (start > 0 && characters.get(i).getX() - characters.get(start - 1).getX() < searchDistance) {
start--;
candidates[neighborInsertionIndex] = new Neighbor(characters.get(start), characters.get(i));
neighborCount++;
if (neighborCount > maxNeighborCount) {
neighborInsertionIndex = clearMostDistant(candidates);
neighborCount--;
} else {
neighborInsertionIndex++;
}
newCandidatesFound = true;
}
while (end < characters.size() && characters.get(end).getX() - characters.get(i).getX() < searchDistance) {
candidates[neighborInsertionIndex] = new Neighbor(characters.get(end), characters.get(i));
neighborCount++;
if (neighborCount > maxNeighborCount) {
neighborInsertionIndex = clearMostDistant(candidates);
neighborCount--;
} else {
neighborInsertionIndex++;
}
end++;
newCandidatesFound = true;
}
if (newCandidatesFound && neighborCount >= maxNeighborCount) {
distance = maxDistance(candidates);
}
}
if (neighborCount < maxNeighborCount) {
clearMostDistant(candidates);
}
List<Neighbor> candidatesList = new ArrayList<>(maxNeighborCount);
for (Neighbor candidate : candidates) {
if (candidate != null) {
candidatesList.add(candidate);
}
}
candidatesList.sort(Comparator.comparingDouble(Neighbor::getDistance));
assert candidatesList.size() == maxNeighborCount;
characters.get(i).setNeighbors(candidatesList);
}
}
private double maxDistance(Neighbor[] candidates) {
double maxDistance = 0;
for (Neighbor candidate : candidates) {
if (candidate == null) {
continue;
}
if (candidate.getDistance() > maxDistance) {
maxDistance = candidate.getDistance();
}
}
return maxDistance;
}
private int clearMostDistant(Neighbor[] candidates) {
double maxDistance = 0;
int maxIndex = 0;
for (int i = 0; i < candidates.length; i++) {
Neighbor candidate = candidates[i];
if (candidate == null) {
continue;
}
if (candidate.getDistance() > maxDistance) {
maxDistance = candidate.getDistance();
maxIndex = i;
}
}
candidates[maxIndex] = null;
return maxIndex;
}
}
@@ -65,7 +65,7 @@ public class ZoneBuilderService {
List<Zone> zones = new ArrayList<>();
unionFind.getGroups().forEach(group -> {
zones.add(mergeLinesInZone(new ArrayList<>(group), characterSpacing, lineSpacing));
zones.add(new Zone(new ArrayList<>(group)));
});
if (zones.size() > MAX_ZONES) {
@@ -208,12 +208,6 @@ public class TextPositionSequence implements CharSequence {
}
public float getTextHeightNoPadding() {
return textPositions.get(0).getHeightDir();
}
public float getTextHeight() {
return textPositions.get(0).getHeightDir() + HEIGHT_PADDING;
@@ -14,7 +14,6 @@ import org.springframework.stereotype.Service;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.DocstrumSegmentationService;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.model.Zone;
import com.knecon.fforesight.service.layoutparser.processor.docstrum.utils.DoubleUtils;
import com.knecon.fforesight.service.layoutparser.processor.model.AbstractPageBlock;
import com.knecon.fforesight.service.layoutparser.processor.model.ClassificationPage;
import com.knecon.fforesight.service.layoutparser.processor.model.FloatFrequencyCounter;
@@ -24,7 +23,9 @@ import com.knecon.fforesight.service.layoutparser.processor.model.table.TablePag
import com.knecon.fforesight.service.layoutparser.processor.model.text.StringFrequencyCounter;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPageBlock;
import com.knecon.fforesight.service.layoutparser.processor.model.text.TextPositionSequence;
import com.knecon.fforesight.service.layoutparser.processor.utils.QuickSort;
import com.knecon.fforesight.service.layoutparser.processor.utils.RulingTextDirAdjustUtil;
import com.knecon.fforesight.service.layoutparser.processor.utils.TextPositionSequenceComparator;
import lombok.RequiredArgsConstructor;
@@ -52,17 +53,13 @@ public class DocstrumBlockificationService {
});
var zones = docstrumSegmentationService.segmentPage(textPositions, xyOrder);
var pageBlocks = toAbstractPageBlocks(zones, usedHorizonalRulings, usedVerticalRulings, xyOrder);
var pageBlocks = toAbstractPageBlocks(zones, usedHorizonalRulings, usedVerticalRulings);
var classificationPage = new ClassificationPage(pageBlocks);
mergeIntersectingBlocks(classificationPage.getTextBlocks(), 0, 0);
return classificationPage;
return new ClassificationPage(pageBlocks);
}
private List<AbstractPageBlock> toAbstractPageBlocks(List<Zone> zones, List<Ruling> horizontalRulings, List<Ruling> verticalRulings, boolean xyOrder) {
private List<AbstractPageBlock> toAbstractPageBlocks(List<Zone> zones, List<Ruling> horizontalRulings, List<Ruling> verticalRulings) {
List<AbstractPageBlock> abstractPageBlocks = new ArrayList<>();
zones.forEach(zone -> {
@@ -77,31 +74,20 @@ public class DocstrumBlockificationService {
abstractPageBlocks.addAll(splitZonesAtRulings(textPositionSequences, horizontalRulings, verticalRulings));
});
if (xyOrder) {
abstractPageBlocks.sort(Comparator.comparing(AbstractPageBlock::getMinY, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD))
.thenComparing(AbstractPageBlock::getMinX, (o1, o2) -> DoubleUtils.compareDouble(o1, o2, THRESHOLD)));
abstractPageBlocks.sort(new Comparator<AbstractPageBlock>() {
@Override
public int compare(AbstractPageBlock o1, AbstractPageBlock o2) {
return Math.abs(o1.getMinY() - o2.getMinY()) < 5 && o1.getMinX() < o2.getMinX() == true ? -1 : 0;
}
});
}
return abstractPageBlocks;
}
public void combineBlocks(ClassificationPage page) {
mergeIntersectingBlocks(page.getTextBlocks(), 0, 0);
TextPageBlock previous = new TextPageBlock();
ListIterator<AbstractPageBlock> itty = page.getTextBlocks().listIterator();
while (itty.hasNext()) {
AbstractPageBlock block = itty.next();
if (block instanceof TablePageBlock) {
previous = new TextPageBlock();
continue;
}
TextPageBlock current = (TextPageBlock) block;
@@ -114,22 +100,22 @@ public class DocstrumBlockificationService {
}
if (areTheOnlyTwoBlocksOnHeightsWithBothMoreThanTwoLines(previous, current, page)) {
previous = combineBlocksAndResetIterator(previous, current, itty, true);
combineBlocksAndResetIterator(previous, current, itty, true);
continue;
}
if (previous.almostIntersects(current, 0, 0)) {
previous = combineBlocksAndResetIterator(previous, current, itty, previous.isToDuplicate());
combineBlocksAndResetIterator(previous, current, itty, previous.isToDuplicate());
continue;
}
if (isSameTopOrBottomWithPreviousHavingMaxTwoLinesAndCurrentThanOneAndMax4OtherBlocksOnHeight(previous, current, page)) {
previous = combineBlocksAndResetIterator(previous, current, itty, previous.isToDuplicate());
combineBlocksAndResetIterator(previous, current, itty, false);
continue;
}
if (isOnlyIntersectingYAndOnelineOrPrevoiusTwoLines(previous, current, page)) {
previous = combineBlocksAndResetIterator(previous, current, itty, previous.isToDuplicate());
combineBlocksAndResetIterator(previous, current, itty, false);
continue;
}
@@ -137,7 +123,7 @@ public class DocstrumBlockificationService {
previous = current;
}
mergeIntersectingBlocks(page.getTextBlocks(), 0, 6.5f);
mergeIntersectingBlocks(page.getTextBlocks(), 0, 0);
}
@@ -154,7 +140,7 @@ public class DocstrumBlockificationService {
ClassificationPage page) {
return (Math.abs(previous.getMaxY() - current.getMaxY()) < THRESHOLD || Math.abs(previous.getMinY() - current.getMinY()) < THRESHOLD) //
&& (previous.getNumberOfLines() == 1 && current.getNumberOfLines() >= 1 || previous.getNumberOfLines() >= 2 && current.getNumberOfLines() == 1) //
&& (previous.getNumberOfLines() == 1 && current.getNumberOfLines() >= 1 || previous.getNumberOfLines() == 2 && current.getNumberOfLines() == 1) //
&& !hasBetween(current, previous, page.getTextBlocks()) && numberOfYIntersectionsOfSmallerBlocksWithOtherBlocks(current, previous, page.getTextBlocks()) <= 4;
}
@@ -167,7 +153,7 @@ public class DocstrumBlockificationService {
}
private TextPageBlock combineBlocksAndResetIterator(TextPageBlock previous, TextPageBlock current, ListIterator<AbstractPageBlock> itty, boolean toDuplicate) {
private void combineBlocksAndResetIterator(TextPageBlock previous, TextPageBlock current, ListIterator<AbstractPageBlock> itty, boolean toDuplicate) {
previous.getSequences().addAll(current.getSequences());
previous = buildTextBlock(previous.getSequences(), 0);
@@ -176,7 +162,6 @@ public class DocstrumBlockificationService {
itty.previous();
itty.set(previous);
itty.next();
return previous;
}
@@ -221,20 +206,22 @@ public class DocstrumBlockificationService {
public void mergeIntersectingBlocks(List<AbstractPageBlock> blocks, float xThreshold, float yThreshold) {
ListIterator<AbstractPageBlock> itty = blocks.listIterator();
Set<AbstractPageBlock> toRemove = new HashSet<>();
while (itty.hasNext()) {
AbstractPageBlock block = itty.next();
if(block == null){
continue;
}
if (block instanceof TablePageBlock) {
continue;
}
TextPageBlock current = (TextPageBlock) block;
if (current.isToDuplicate()) {
continue;
}
for (int i = 0; i < blocks.size(); i++) {
if(blocks.get(i) == null){
if (toRemove.contains(blocks.get(i))) {
continue;
}
if (blocks.get(i) == current) {
@@ -246,24 +233,21 @@ public class DocstrumBlockificationService {
TextPageBlock inner = (TextPageBlock) blocks.get(i);
if (inner.isToDuplicate()) {
continue;
}
if (current.getDir() == inner.getDir() && current.almostIntersects(inner, yThreshold, xThreshold)) {
boolean toDuplicate = current.isToDuplicate() || inner.isToDuplicate();
current.getSequences().addAll(inner.getSequences());
QuickSort.sort(current.getSequences(), new TextPositionSequenceComparator());
current = buildTextBlock(current.getSequences(), 0);
current.setToDuplicate(toDuplicate);
blocks.set(i, null);
toRemove.add(inner);
itty.set(current);
}
}
}
var blocksIterator = blocks.iterator();
while(blocksIterator.hasNext()){
if(blocksIterator.next() == null){
blocksIterator.remove();
}
}
blocks.removeAll(toRemove);
}
@@ -91,7 +91,7 @@ public class SectionNodeFactory {
if (abstractPageBlock instanceof TextPageBlock) {
switch (layoutParsingType) {
case REDACT_MANAGER, CLARIFYND, CLARIFYND_PARAGRAPH_DEBUG -> {
case CLARIFYND, CLARIFYND_PARAGRAPH_DEBUG -> {
alreadyMerged.add(abstractPageBlock);
remainingBlocks.remove(abstractPageBlock);
DocumentGraphFactory.addParagraphOrHeadline(section, (TextPageBlock) abstractPageBlock, context, new ArrayList<>());
@@ -28,13 +28,15 @@ 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;
}
@@ -46,19 +48,25 @@ 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.getTextHeightNoPadding();
float pos2YTop = pos2YBottom - pos2.getTextHeightNoPadding();
float pos1YTop = pos1YBottom - pos1.getTextHeight();
float pos2YTop = pos2YBottom - pos2.getTextHeight();
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;
}
}
}
@@ -27,7 +27,7 @@ public class ViewerDocumentTest extends BuildDocumentTest {
@SneakyThrows
public void testViewerDocument() {
String fileName = "files/new/ScrambledTextAfterSorting.pdf";
String fileName = "files/bdr/notMergedParagraphs.pdf";
String tmpFileName = "/tmp/" + Path.of(fileName).getFileName() + "_VIEWER.pdf";
var documentFile = new ClassPathResource(fileName).getFile();
@@ -35,7 +35,7 @@ public class ViewerDocumentTest extends BuildDocumentTest {
LayoutGridService layoutGridService = new LayoutGridService(viewerDocumentService);
long start = System.currentTimeMillis();
Document document = buildGraph(fileName, LayoutParsingType.DOCUMINE);
Document document = buildGraph(fileName, LayoutParsingType.CLARIFYND);
layoutGridService.addLayoutGrid(documentFile, document, new File(tmpFileName), true);
System.out.printf("Total time: %.2fs%n", ((float) (System.currentTimeMillis() - start)) / 1000);
}
@@ -59,7 +59,7 @@ public class ViewerDocumentTest extends BuildDocumentTest {
new ImageServiceResponse(),
tableResponse,
new VisualLayoutParsingResponse(),
Map.of("file", Path.of(fileName).getFileName().toFile().toString()));
Map.of("file",Path.of(fileName).getFileName().toFile().toString()));
ViewerDocumentService viewerDocumentService = new ViewerDocumentService(null);
LayoutGridService layoutGridService = new LayoutGridService(viewerDocumentService);
Document document = DocumentGraphFactory.buildDocumentGraph(LayoutParsingType.DOCUMINE, classificationDocument);