ADFA-3813 | Fix OCR image metadata parsing and refactor CV domain

cv-image-to-xml/src/main/java/org/appdevforall/codeonthego/computervision/data/repository/VisionRepository.kt

@@ -0,0 +1,18 @@
+package org.appdevforall.codeonthego.computervision.data.repository
+
+import android.graphics.Bitmap
+import com.google.mlkit.vision.text.Text
+import org.appdevforall.codeonthego.computervision.domain.model.DetectionResult
+
+/**
+ * Abstract contract for computer vision data sources.
+ * Handles raw interactions with machine learning models (YOLO, MLKit)
+ * without leaking domain logic.
+ */
+interface VisionRepository {
+    suspend fun initModel(): Result<Unit>
+    suspend fun detectWidgets(bitmap: Bitmap): Result<List<DetectionResult>>
+    suspend fun recognizeText(bitmap: Bitmap): Result<List<Text.TextBlock>>
+    fun isInitialized(): Boolean
+    fun release()
+}

cv-image-to-xml/src/main/java/org/appdevforall/codeonthego/computervision/data/repository/VisionRepositoryImpl.kt

@@ -0,0 +1,39 @@
+package org.appdevforall.codeonthego.computervision.data.repository
+
+import android.content.res.AssetManager
+import android.graphics.Bitmap
+import com.google.mlkit.vision.text.Text
+import kotlinx.coroutines.Dispatchers
+import kotlinx.coroutines.withContext
+import org.appdevforall.codeonthego.computervision.data.source.OcrSource
+import org.appdevforall.codeonthego.computervision.data.source.YoloModelSource
+import org.appdevforall.codeonthego.computervision.domain.model.DetectionResult
+
+class VisionRepositoryImpl(
+    private val assetManager: AssetManager,
+    private val yoloModelSource: YoloModelSource,
+    private val ocrSource: OcrSource
+) : VisionRepository {
+
+    override suspend fun initModel(): Result<Unit> = withContext(Dispatchers.IO) {
+        runCatching {
+            yoloModelSource.initialize(assetManager)
+        }
+    }
+
+    override suspend fun detectWidgets(bitmap: Bitmap): Result<List<DetectionResult>> =
+        withContext(Dispatchers.Default) {
+            runCatching { yoloModelSource.runInference(bitmap) }
+        }
+
+    override suspend fun recognizeText(bitmap: Bitmap): Result<List<Text.TextBlock>> =
+        withContext(Dispatchers.Default) {
+            ocrSource.recognizeText(bitmap)
+        }
+
+    override fun isInitialized(): Boolean = yoloModelSource.isInitialized()
+
+    override fun release() {
+        yoloModelSource.release()
+    }
+}

cv-image-to-xml/src/main/java/org/appdevforall/codeonthego/computervision/di/ComputerVisionModule.kt

@@ -1,11 +1,17 @@
 package org.appdevforall.codeonthego.computervision.di
 
-import org.appdevforall.codeonthego.computervision.data.repository.ComputerVisionRepository
-import org.appdevforall.codeonthego.computervision.data.repository.ComputerVisionRepositoryImpl
 import org.appdevforall.codeonthego.computervision.data.repository.DrawableImportHelper
+import org.appdevforall.codeonthego.computervision.data.repository.VisionRepository
+import org.appdevforall.codeonthego.computervision.data.repository.VisionRepositoryImpl
 import org.appdevforall.codeonthego.computervision.data.source.OcrSource
 import org.appdevforall.codeonthego.computervision.data.source.YoloModelSource
+import org.appdevforall.codeonthego.computervision.domain.GenericBoxResolver
 import org.appdevforall.codeonthego.computervision.domain.RegionOcrProcessor
+import org.appdevforall.codeonthego.computervision.domain.usecase.GenerateXmlUC
+import org.appdevforall.codeonthego.computervision.domain.usecase.ImportPlaceholderImageUC
+import org.appdevforall.codeonthego.computervision.domain.usecase.PrepareImageUC
+import org.appdevforall.codeonthego.computervision.domain.usecase.RemovePlaceholderImageUC
+import org.appdevforall.codeonthego.computervision.domain.usecase.RunVisionUC
 import org.appdevforall.codeonthego.computervision.ui.viewmodel.ComputerVisionViewModel
 import org.koin.android.ext.koin.androidContext
 import org.koin.core.module.dsl.viewModel
@@ -14,30 +20,33 @@ import org.koin.dsl.module
 val computerVisionModule = module {
 
     single { YoloModelSource() }
-
     single { OcrSource() }
-
     single { RegionOcrProcessor(ocrSource = get()) }
+    single { GenericBoxResolver() }
 
-    single<ComputerVisionRepository> {
-        ComputerVisionRepositoryImpl(
+    single<VisionRepository> {
+        VisionRepositoryImpl(
             assetManager = androidContext().assets,
             yoloModelSource = get(),
-            regionOcrProcessor = get()
+            ocrSource = get()
         )
     }
 
-    single {
-        DrawableImportHelper(
-            contentResolver = androidContext().contentResolver
-        )
-    }
+    single { DrawableImportHelper(contentResolver = androidContext().contentResolver) }
+    single { GenerateXmlUC() }
+    single { ImportPlaceholderImageUC(drawableImportHelper = get()) }
+    single { PrepareImageUC(contentResolver = androidContext().contentResolver) }
+    single { RemovePlaceholderImageUC(drawableImportHelper = get()) }
+    single { RunVisionUC(repository = get(), boxResolver = get(), regionOcrProcessor = get()) }
 
     viewModel { (layoutFilePath: String?, layoutFileName: String?) ->
         ComputerVisionViewModel(
             repository = get(),
-            drawableImportHelper = get(),
-            contentResolver = androidContext().contentResolver,
+            prepareImageUC = get(),
+            runVisionUC = get(),
+            generateXmlUC = get(),
+            importPlaceholderImageUC = get(),
+            removePlaceholderImageUC = get(),
             layoutFilePath = layoutFilePath,
             layoutFileName = layoutFileName
         )

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

@@ -0,0 +1,38 @@
+package org.appdevforall.codeonthego.computervision.domain
+
+import android.graphics.Rect
+import org.appdevforall.codeonthego.computervision.domain.model.DetectionResult
+import org.appdevforall.codeonthego.computervision.domain.model.ScaledBox
+import kotlin.math.max
+import kotlin.math.roundToInt
+
+/**
+ * Scales the normalized YOLO coordinates (0.0 to 1.0) to the target dimensions
+ * in DP (e.g., 360x640) of the Android screen.
+ */
+object DetectionScaler {
+    private const val MIN_W_ANY = 8
+    private const val MIN_H_ANY = 8
+
+    fun scale(
+        detection: DetectionResult, sourceWidth: Int, sourceHeight: Int, targetW: Int, targetH: Int
+    ): ScaledBox {
+        if (sourceWidth == 0 || sourceHeight == 0) {
+            return ScaledBox(detection.label, detection.text, 0, 0, MIN_W_ANY, MIN_H_ANY, MIN_W_ANY / 2, MIN_H_ANY / 2, Rect(0, 0, MIN_W_ANY, MIN_H_ANY))
+        }
+        val rect = detection.boundingBox
+        val normCx = ((rect.left + rect.right) / 2f) / sourceWidth.toFloat()
+        val normCy = ((rect.top + rect.bottom) / 2f) / sourceHeight.toFloat()
+        val normW = (rect.right - rect.left) / sourceWidth.toFloat()
+        val normH = (rect.bottom - rect.top) / sourceHeight.toFloat()
+
+        val x = max(0, ((normCx - normW / 2f) * targetW).roundToInt())
+        val y = max(0, ((normCy - normH / 2f) * targetH).roundToInt())
+        val w = max(MIN_W_ANY, (normW * targetW).roundToInt())
+        val h = max(MIN_H_ANY, (normH * targetH).roundToInt())
+
+        return ScaledBox(
+            detection.label, detection.text, x, y, w, h, x + w / 2, y + h / 2, Rect(x, y, x + w, y + h)
+        )
+    }
+}