ADFA-3757 | Fix fuzzy attribute parsing and OCR text assembly

cv-image-to-xml/src/main/java/org/appdevforall/codeonthego/computervision/domain/FuzzyAttributeParser.kt

@@ -26,8 +26,8 @@ object FuzzyAttributeParser {
  TEXT("android:text", listOf("text")),
  HINT("android:hint", listOf("hint")),
  BACKGROUND("android:background", listOf("background", "bg"), ValueType.COLOR),
- BACKGROUND_TINT("app:backgroundTint", listOf("backgroundtint", "background_tint"), ValueType.COLOR),
- SRC("android:src", listOf("src", "scr"), ValueType.DRAWABLE),
+ BACKGROUND_TINT("app:backgroundTint", listOf("backgroundtint", "background_tint", "bg_tint"), ValueType.COLOR),
+ SRC("android:src", listOf("src", "scr", "sre", "5rc"), ValueType.DRAWABLE),
  CONTENT_DESCRIPTION("android:contentDescription", listOf("contentdescription", "content_description")),
 
  TEXT_SIZE("android:textSize", listOf("textsize", "text_size"), ValueType.SP_DIMENSION),
@@ -74,7 +74,7 @@ object FuzzyAttributeParser {
  LAYOUT_MARGIN_RIGHT("android:layout_marginRight", listOf("layout_marginright", "layout_margin_right", "margin_right"), ValueType.DIMENSION),
 
  LAYOUT_WEIGHT("android:layout_weight", listOf("layout_weight", "weight"), ValueType.FLOAT),
- LAYOUT_GRAVITY("android:layout_gravity", listOf("layout_gravity")),
+ LAYOUT_GRAVITY("android:layout_gravity", listOf("layout_gravity", "layaut_gravity")),
  GRAVITY("android:gravity", listOf("gravity")),
  ORIENTATION("android:orientation", listOf("orientation")),
 
@@ -150,6 +150,7 @@ object FuzzyAttributeParser {
  private val ocrLetterZToTwoRegex = Regex("[zZ]")
  private val ocrLetterSToFiveRegex = Regex("[sS]")
  private val ocrLetterBToSixRegex = Regex("[bB]")
+ private val ocrDenoiseRegex = Regex("inm|rn|wm|nm")
 
  private val matchKeywords = setOf("match", "parent")
  private val wrapKeywords = setOf("wrap", "content", "wrapcan")
@@ -419,7 +420,7 @@ object FuzzyAttributeParser {
  .replace(Regex("op$"), "dp")
  .replace(Regex("olp$"), "dp")
 
- val numericString = fixedUnit.replace(Regex("[a-z]+$"), "")
+ val numericString = fixedUnit.replace("_", "")
  val numericPart = extractOcrNumber(numericString)
 
  if (numericPart != null) return "${numericPart}dp"
@@ -432,7 +433,7 @@ object FuzzyAttributeParser {
  .replace(" ", "")
  .replace(Regex("5p$"), "sp")
 
- val numericString = fixedUnit.replace(Regex("[a-z]+$"), "")
+ val numericString = fixedUnit.replace("_", "")
  val numericPart = extractOcrNumber(numericString)
 
  if (numericPart != null) return "${numericPart}sp"
@@ -457,22 +458,30 @@ object FuzzyAttributeParser {
  }
 
  private fun cleanId(value: String): String {
- return value.lowercase()
+ return denoiseOcrIdentifier(value.lowercase())
  .replace(nonAlphanumericRegex, "_")
  .replace(multipleUnderscoresRegex, "_")
  .trimEnd('_')
  .trimStart('_')
  }
 
- private fun denoiseOcrIdentifier(value: String): String =
- value.replace("rn", "m")
- .replace("wm", "m")
+ private fun denoiseOcrIdentifier(value: String): String {
+ return value.replace(ocrDenoiseRegex) { matchResult ->
+ when (matchResult.value) {
+ "inm" -> "im"
+ else -> "m"
+ }
+ }
+ }
 
  private fun cleanDrawable(value: String): String {
  if (value.startsWith("@drawable/")) return value
- val cleaned = value.replace(Regex("[^a-zA-Z0-9_.]"), "")
- .substringBeforeLast('.')
- .lowercase()
+
+ var cleaned = value.lowercase()
+ .replace(Regex("\\.(png|jpg|jpeg|webp|xml|svg)$"), "")
+
+ cleaned = cleaned.replace(Regex("[^a-z0-9_]"), "")
+
  return "@drawable/${denoiseOcrIdentifier(cleaned)}"
  }
 

cv-image-to-xml/src/main/java/org/appdevforall/codeonthego/computervision/domain/RegionOcrProcessor.kt

@@ -9,6 +9,7 @@ import kotlinx.coroutines.coroutineScope
 import org.appdevforall.codeonthego.computervision.data.source.OcrSource
 import org.appdevforall.codeonthego.computervision.domain.model.DetectionResult
 import org.appdevforall.codeonthego.computervision.utils.BitmapUtils
+import org.appdevforall.codeonthego.computervision.utils.OcrTextAssembler
 
 class RegionOcrProcessor(
  private val ocrSource: OcrSource,
@@ -69,12 +70,8 @@ class RegionOcrProcessor(
  try {
  crop = BitmapUtils.cropRegion(bitmap, component.boundingBox, componentPadding)
  preprocessed = BitmapUtils.preprocessForOcr(crop)
- val text = ocrSource.recognizeText(preprocessed)
- .getOrNull()
- ?.joinToString(" ") { it.text }
- ?.replace("\n", " ")
- ?.trim()
- ?: ""
+ val textBlocks = ocrSource.recognizeText(preprocessed).getOrNull()
+ val text = textBlocks?.let { OcrTextAssembler.extractTextWithTolerance(it) } ?: ""
  component.copy(text = text)
  } finally {
  preprocessed?.recycle()
@@ -124,7 +121,7 @@ class RegionOcrProcessor(
  ),
  label = "text",
  score = 0.99f,
- text = line.text.trim(),
+ text = OcrTextAssembler.joinElementsWithTolerance(line),
  isYolo = false
  )
  }

cv-image-to-xml/src/main/java/org/appdevforall/codeonthego/computervision/utils/OcrTextAssembler.kt

@@ -0,0 +1,44 @@
+package org.appdevforall.codeonthego.computervision.utils
+
+import com.google.mlkit.vision.text.Text
+
+object OcrTextAssembler {
+ const val DEFAULT_SPACE_GAP_TOLERANCE_PX = 15f
+
+ fun extractTextWithTolerance(
+ textBlocks: List<Text.TextBlock>,
+ maxSpaceGap: Float = DEFAULT_SPACE_GAP_TOLERANCE_PX
+ ): String {
+ return textBlocks.joinToString(" ") { block ->
+ block.lines.joinToString(" ") { line ->
+ joinElementsWithTolerance(line, maxSpaceGap)
+ }
+ }.trim()
+ }
+
+ fun joinElementsWithTolerance(line: Text.Line, maxSpaceGap: Float = DEFAULT_SPACE_GAP_TOLERANCE_PX): String {
+ val elements = line.elements.sortedBy { it.boundingBox?.left ?: 0 }
+ if (elements.isEmpty()) return ""
+
+ val builder = StringBuilder()
+ var prevRight = elements.first().boundingBox?.right ?: 0
+ builder.append(elements.first().text)
+
+ for (i in 1 until elements.size) {
+ val current = elements[i]
+ val currentBox = current.boundingBox
+
+ if (currentBox != null) {
+ val gap = currentBox.left - prevRight
+ if (gap > maxSpaceGap) {
+ builder.append(" ")
+ }
+ prevRight = currentBox.right
+ } else {
+ builder.append(" ")
+ }
+ builder.append(current.text)
+ }
+ return builder.toString().trim()
+ }
+}