Removed example tests, corrected private variable naming

Author: ThomasKiljanczykDev

android/app/detector/src/androidTest/java/org/tensorflow/lite/examples/detector/ExampleInstrumentedTest.kt

@@ -39,20 +39,20 @@ open class MainActivity : AppCompatActivity() {
     private lateinit var mDetector: Detector
     private var mDetectionProcessor: DetectionProcessor? = null
 
-    private lateinit var sourceBitmap: Bitmap
-    private lateinit var cropBitmap: Bitmap
+    private lateinit var mSourceBitmap: Bitmap
+    private lateinit var mCropBitmap: Bitmap
 
     override fun onCreate(savedInstanceState: Bundle?) {
         super.onCreate(savedInstanceState)
         mBinding = ActivityMainBinding.inflate(layoutInflater)
         setContentView(mBinding.root)
 
-        sourceBitmap = assets.open("kite.jpg").use { inputStream ->
+        mSourceBitmap = assets.open("kite.jpg").use { inputStream ->
             BitmapFactory.decodeStream(inputStream)
         }
 
-        cropBitmap = processBitmap(sourceBitmap, DETECTION_MODEL.inputSize)
-        mBinding.imageView.setImageBitmap(cropBitmap)
+        mCropBitmap = processBitmap(mSourceBitmap, DETECTION_MODEL.inputSize)
+        mBinding.imageView.setImageBitmap(mCropBitmap)
 
         setUpDetector()
         lifecycleScope.launch(Dispatchers.Main) {
@@ -104,7 +104,7 @@ open class MainActivity : AppCompatActivity() {
 
         mBinding.detectButton.setOnClickListener {
             lifecycleScope.launch(Dispatchers.Default) {
-                mDetectionProcessor?.processImage(cropBitmap)
+                mDetectionProcessor?.processImage(mCropBitmap)
             }
         }
     }

android/app/detector/src/test/java/org/tensorflow/lite/examples/detector/ExampleUnitTest.kt

@@ -57,7 +57,7 @@ class DetectionProcessor(
             previewWidth,
             previewHeight,
             ((rotation + 1) % 4) * 90,
-            showScore = SHOW_SCORE
+            mShowScore = SHOW_SCORE
         )
         trackingOverlay.setTracker(mTracker)
     }

android/app/src/main/java/org/tensorflow/lite/examples/detector/activities/MainActivity.kt

@@ -20,7 +20,7 @@ interface Detector {
          * A unique identifier for what has been recognized. Specific to the detected class,
          * not the instance of the [Detection] object.
          */
-        private val id: String,
+        private val mId: String,
         /**
          * Display name for the [Detection].
          */
@@ -36,7 +36,7 @@ interface Detector {
         val detectedClass: Int
     ) : Comparable<Detection> {
         override fun toString(): String {
-            var resultString = "[$id] $className "
+            var resultString = "[$mId] $className "
             resultString += "(%.1f%%) ".format(score * 100.0f)
             resultString += "$boundingBox"
             return resultString

android/app/src/main/java/org/tensorflow/lite/examples/detector/helpers/DetectionProcessor.kt

@@ -21,8 +21,8 @@ import kotlin.math.min
 @Suppress("UNNECESSARY_NOT_NULL_ASSERTION")
 internal class YoloV4Detector(
     assetManager: AssetManager,
-    private val detectionModel: DetectionModel,
-    private val minimumScore: Float,
+    private val mDetectionModel: DetectionModel,
+    private val mMinimumScore: Float,
 ) : Detector {
 
     private companion object {
@@ -33,53 +33,50 @@ internal class YoloV4Detector(
 
     }
 
-    private val inputSize: Int = detectionModel.inputSize
+    private val mInputSize: Int = mDetectionModel.inputSize
 
     // Config values.
-    // Pre-allocated buffers.
-    private val labels: List<String>
-    private val interpreter: Interpreter
+    private val mLabels: List<String>
+    private val mInterpreter: Interpreter
     private val mNmsThresh = 0.6f
 
-    private val intValues = IntArray(inputSize * inputSize)
-
-    private val byteBuffer: Array<ByteBuffer>
-
-    private val outputMap: MutableMap<Int, Array<Array<FloatArray>>> = HashMap()
+    // Pre-allocated buffers.
+    private val intValues = IntArray(mInputSize * mInputSize)
+    private val mByteBuffer: Array<ByteBuffer>
+    private val mOutputMap: MutableMap<Int, Array<Array<FloatArray>>> = HashMap()
 
     init {
-
-        val labelsFilename = detectionModel.labelFilePath
+        val labelsFilename = mDetectionModel.labelFilePath
             .split("file:///android_asset/")
             .toTypedArray()[1]
 
-        labels = assetManager.open(labelsFilename)
+        mLabels = assetManager.open(labelsFilename)
             .use { it.readBytes() }
             .decodeToString()
             .trim()
             .split("\n")
             .map { it.trim() }
 
-        interpreter = initializeInterpreter(assetManager)
+        mInterpreter = initializeInterpreter(assetManager)
 
-        val numBytesPerChannel = if (detectionModel.isQuantized) {
+        val numBytesPerChannel = if (mDetectionModel.isQuantized) {
             1 // Quantized (int8)
         } else {
             4 // Floating point (fp32)
         }
 
         // input size * input size * pixel count (RGB) * pixel size (int8/fp32)
-        byteBuffer = arrayOf(
-            ByteBuffer.allocateDirect(inputSize * inputSize * 3 * numBytesPerChannel)
+        mByteBuffer = arrayOf(
+            ByteBuffer.allocateDirect(mInputSize * mInputSize * 3 * numBytesPerChannel)
         )
-        byteBuffer[0].order(ByteOrder.nativeOrder())
+        mByteBuffer[0].order(ByteOrder.nativeOrder())
 
-        outputMap[0] = arrayOf(Array(detectionModel.outputSize) { FloatArray(numBytesPerChannel) })
-        outputMap[1] = arrayOf(Array(detectionModel.outputSize) { FloatArray(labels.size) })
+        mOutputMap[0] = arrayOf(Array(mDetectionModel.outputSize) { FloatArray(numBytesPerChannel) })
+        mOutputMap[1] = arrayOf(Array(mDetectionModel.outputSize) { FloatArray(mLabels.size) })
     }
 
     override fun getDetectionModel(): DetectionModel {
-        return detectionModel
+        return mDetectionModel
     }
 
     override fun runDetection(bitmap: Bitmap): List<Detection> {
@@ -105,7 +102,7 @@ internal class YoloV4Detector(
             }
         }
 
-        return assetManager.openFd(detectionModel.modelFilename).use { fileDescriptor ->
+        return assetManager.openFd(mDetectionModel.modelFilename).use { fileDescriptor ->
             val fileInputStream = FileInputStream(fileDescriptor.fileDescriptor)
             val fileByteBuffer = fileInputStream.channel.map(
                 FileChannel.MapMode.READ_ONLY,
@@ -122,36 +119,36 @@ internal class YoloV4Detector(
      */
     private fun convertBitmapToByteBuffer(bitmap: Bitmap) {
         val startTime = SystemClock.uptimeMillis()
-        val scaledBitmap = Bitmap.createScaledBitmap(bitmap, inputSize, inputSize, true)
+        val scaledBitmap = Bitmap.createScaledBitmap(bitmap, mInputSize, mInputSize, true)
 
-        scaledBitmap.getPixels(intValues, 0, inputSize, 0, 0, inputSize, inputSize)
+        scaledBitmap.getPixels(intValues, 0, mInputSize, 0, 0, mInputSize, mInputSize)
         scaledBitmap.recycle()
 
-        byteBuffer[0].clear()
+        mByteBuffer[0].clear()
         for (pixel in intValues) {
             val r = (pixel and 0xFF) / 255.0f
             val g = (pixel shr 8 and 0xFF) / 255.0f
             val b = (pixel shr 16 and 0xFF) / 255.0f
 
-            byteBuffer[0].putFloat(r)
-            byteBuffer[0].putFloat(g)
-            byteBuffer[0].putFloat(b)
+            mByteBuffer[0].putFloat(r)
+            mByteBuffer[0].putFloat(g)
+            mByteBuffer[0].putFloat(b)
         }
         Log.v(TAG, "ByteBuffer conversion time : ${SystemClock.uptimeMillis() - startTime} ms")
     }
 
     private fun getDetections(imageWidth: Int, imageHeight: Int): List<Detection> {
-        interpreter.runForMultipleInputsOutputs(byteBuffer, outputMap as Map<Int, Any>)
+        mInterpreter.runForMultipleInputsOutputs(mByteBuffer, mOutputMap as Map<Int, Any>)
 
-        val boundingBoxes = outputMap[0]!![0]
-        val outScore = outputMap[1]!![0]
+        val boundingBoxes = mOutputMap[0]!![0]
+        val outScore = mOutputMap[1]!![0]
 
         return outScore.zip(boundingBoxes)
             .mapIndexedNotNull { index, (classScores, boundingBoxes) ->
-                val bestClassIndex: Int = labels.indices.maxByOrNull { classScores[it] }!!
+                val bestClassIndex: Int = mLabels.indices.maxByOrNull { classScores[it] }!!
                 val bestScore = classScores[bestClassIndex]
 
-                if (bestScore <= minimumScore) {
+                if (bestScore <= mMinimumScore) {
                     return@mapIndexedNotNull null
                 }
 
@@ -167,8 +164,8 @@ internal class YoloV4Detector(
                 )
 
                 return@mapIndexedNotNull Detection(
-                    id = index.toString(),
-                    className = labels[bestClassIndex],
+                    mId = index.toString(),
+                    className = mLabels[bestClassIndex],
                     detectedClass = bestClassIndex,
                     score = bestScore,
                     boundingBox = rectF
@@ -179,7 +176,7 @@ internal class YoloV4Detector(
     private fun nms(detections: List<Detection>): List<Detection> {
         val nmsList: MutableList<Detection> = mutableListOf()
 
-        for (labelIndex in labels.indices) {
+        for (labelIndex in mLabels.indices) {
             val priorityQueue = PriorityQueue<Detection>(50)
             priorityQueue.addAll(detections.filter { it.detectedClass == labelIndex })
 

android/detector/src/main/java/org/tensorflow/lite/examples/detector/Detector.kt

@@ -11,29 +11,29 @@ import java.nio.ByteBuffer
  * Utility class for converting [Image] to [Bitmap].
  */
 class ImageToBitmapConverter(context: Context, image: Image) {
-    private val bitmap: Bitmap =
+    private val mBitmap: Bitmap =
         Bitmap.createBitmap(image.width, image.height, Bitmap.Config.ARGB_8888)
 
-    private val renderScript: RenderScript = RenderScript.create(context)
+    private val mRenderScript: RenderScript = RenderScript.create(context)
 
-    private val scriptYuvToRgb: ScriptIntrinsicYuvToRGB =
-        ScriptIntrinsicYuvToRGB.create(renderScript, Element.U8_3(renderScript))
+    private val mScriptYuvToRgb: ScriptIntrinsicYuvToRGB =
+        ScriptIntrinsicYuvToRGB.create(mRenderScript, Element.U8_3(mRenderScript))
 
-    private val elemType = Type.Builder(renderScript, Element.YUV(renderScript))
+    private val mElemType = Type.Builder(mRenderScript, Element.YUV(mRenderScript))
         .setYuvFormat(ImageFormat.YUV_420_888)
         .create()
 
-    private val inputAllocation: Allocation =
+    private val mInputAllocation: Allocation =
         Allocation.createSized(
-            renderScript,
-            elemType.element,
+            mRenderScript,
+            mElemType.element,
             image.planes.sumOf { it.buffer.remaining() }
         )
 
-    private val outputAllocation: Allocation = Allocation.createFromBitmap(renderScript, bitmap)
+    private val mOutputAllocation: Allocation = Allocation.createFromBitmap(mRenderScript, mBitmap)
 
     init {
-        scriptYuvToRgb.setInput(inputAllocation)
+        mScriptYuvToRgb.setInput(mInputAllocation)
     }
 
     /**
@@ -42,11 +42,11 @@ class ImageToBitmapConverter(context: Context, image: Image) {
     fun imageToBitmap(image: Image): Bitmap {
         val yuvBuffer: ByteArray = yuv420ToByteArray(image)
 
-        inputAllocation.copyFrom(yuvBuffer)
-        scriptYuvToRgb.forEach(outputAllocation)
-        outputAllocation.copyTo(bitmap)
+        mInputAllocation.copyFrom(yuvBuffer)
+        mScriptYuvToRgb.forEach(mOutputAllocation)
+        mOutputAllocation.copyTo(mBitmap)
 
-        return bitmap
+        return mBitmap
     }
 
     private fun yuv420ToByteArray(image: Image): ByteArray {

android/detector/src/main/java/org/tensorflow/lite/examples/detector/YoloV4Detector.kt

@@ -19,7 +19,7 @@ class MultiBoxTracker(
     private val mFrameWidth: Int,
     private val mFrameHeight: Int,
     private val mOrientation: Int,
-    private val showScore: Boolean = true
+    private val mShowScore: Boolean = true
 ) {
 
     private companion object {
@@ -120,14 +120,14 @@ class MultiBoxTracker(
                 trackedDetection.boxPaint
             )
 
-            val labelString = if (showScore && trackedDetection.title.isNotBlank()) {
+            val labelString = if (mShowScore && trackedDetection.title.isNotBlank()) {
                 "%s %.2f%%".format(
                     trackedDetection.title,
                     100 * trackedDetection.score
                 )
             } else if (trackedDetection.title.isNotBlank()) {
                 trackedDetection.title
-            } else if (showScore) {
+            } else if (mShowScore) {
                 "%.2f%%".format(100 * trackedDetection.score)
             } else ""
 

android/detector/src/main/java/org/tensorflow/lite/examples/detector/utils/ImageToBitmapConverter.kt

@@ -11,7 +11,6 @@ import android.view.View
 class TrackingOverlayView(context: Context, attrs: AttributeSet?) : View(context, attrs) {
     private var mTracker: MultiBoxTracker? = null
 
-
     override fun draw(canvas: Canvas) {
         super.draw(canvas)
         mTracker?.draw(canvas)