Adding isolate

lib/data/remote/models/detection_model.dart

@@ -0,0 +1,22 @@
+import 'package:json_annotation/json_annotation.dart';
+
+part 'detection_model.g.dart';
+
+@JsonSerializable()
+class DetectionModel {
+  List<int> finalClasses;
+  List<List<double>> finalBboxes;
+  List<double> finalScores;
+
+
+  DetectionModel({
+    required this.finalClasses,
+    required this.finalBboxes,
+    required this.finalScores,
+  });
+
+  Map<String, dynamic> toJson() => _$DetectionModelToJson(this);
+
+  factory DetectionModel.fromJson(Map<String, dynamic> json) =>
+      _$DetectionModelFromJson(json);
+}

lib/data/remote/models/recipe_model.g.dart

@@ -17,9 +17,9 @@ RecipeModel _$RecipeModelFromJson(Map<String, dynamic> json) => RecipeModel(
       instructions: (json['instructions'] as List<dynamic>?)
           ?.map((e) => e as String)
           .toList(),
-      prepTime: json['prepTime'] as int?,
-      cookTime: json['cookTime'] as int?,
-      servings: json['servings'] as int?,
+      prepTime: (json['prepTime'] as num?)?.toInt(),
+      cookTime: (json['cookTime'] as num?)?.toInt(),
+      servings: (json['servings'] as num?)?.toInt(),
       utensils: (json['utensils'] as List<dynamic>?)
           ?.map((e) => e as String)
           .toList(),

lib/data/remote/models/utensil_model.g.dart

@@ -7,9 +7,9 @@ part of 'utensil_model.dart';
 // **************************************************************************
 
 Utensil _$UtensilFromJson(Map<String, dynamic> json) => Utensil(
-      id: json['id'] as int?,
+      id: (json['id'] as num?)?.toInt(),
       name: json['name'] as String?,
-      isSelected: json['isSelected'] as int?,
+      isSelected: (json['isSelected'] as num?)?.toInt(),
     );
 
 Map<String, dynamic> _$UtensilToJson(Utensil instance) => <String, dynamic>{

lib/data/remote/responses/base_response.g.dart

@@ -13,7 +13,7 @@ BaseResponse<T> _$BaseResponseFromJson<T>(
     BaseResponse<T>(
       data: _$nullableGenericFromJson(json['data'], fromJsonT),
     )
-      ..code = json['code'] as int?
+      ..code = (json['code'] as num?)?.toInt()
       ..message = json['message'] as String?;
 
 Map<String, dynamic> _$BaseResponseToJson<T>(

lib/presentation/recipe_detection/view_model/recipe_detection_view_model.dart

@@ -65,9 +65,14 @@ class RecipeDetectionViewModel extends BaseViewModel {
     hideLoadingContainer();
   }
 
+  void _resetBounding(){
+    classes.clear();
+    bboxes.clear();
+    scores.clear();
+  }
+
   Future<void> updatePostProcess() async {
     detectedIngredients.clear();
-    print("MASUK 1");
     if (inferenceOutput.isEmpty) {
       return;
     }
@@ -76,8 +81,6 @@ class RecipeDetectionViewModel extends BaseViewModel {
     List<List<double>> newBboxes = [];
     List<double> newScores = [];
 
-    print("MASUK 2");
-
     /// Wait this process with loading
     (newClasses, newBboxes, newScores) = await model.postprocess(
       inferenceOutput,
@@ -86,7 +89,6 @@ class RecipeDetectionViewModel extends BaseViewModel {
       confidenceThreshold: confidenceThreshold,
       iouThreshold: iouThreshold,
     );
-    print("MASUK 3");
 
     debugPrint('Detected ${newClasses} classes');
     debugPrint('Detected ${newBboxes.length} boxed');
@@ -132,17 +134,11 @@ class RecipeDetectionViewModel extends BaseViewModel {
             const CustomCameraWidget(compressionQuality: 80),
       ),
     );
-    print("MASUK 9");
-
 
     if (data != null) {
-      print("MASUK 10");
-
       imageFile.value = data;
       // _startTimer();
-      await Future.delayed(const Duration(milliseconds: 500));
-      print("MASUK 11");
-
+      _resetBounding();
       _detectIngredients();
     }
   }
@@ -152,18 +148,14 @@ class RecipeDetectionViewModel extends BaseViewModel {
   }
 
   void _detectIngredients() async {
-    print("MASUK SINI 1");
-
-    // showLoadingDialog();
-
+    showLoadingDialog();
     final image = img.decodeImage(await imageFile.value!.readAsBytes())!;
-    print("MASUK SINI 2");
 
     imageHeight.value = image.height;
     imageWidth.value = image.width;
-    inferenceOutput.value = model.infer(image);
+    inferenceOutput.value = await model.inferenceImage(image);
+    closeLoadingDialog();
 
-    print("MASUK SINI 3");
     updatePostProcess();
 
 

lib/utils/detection/isolate_inference.dart

@@ -0,0 +1,114 @@
+import 'dart:isolate';
+import 'dart:math';
+
+import 'package:flutter/foundation.dart';
+import 'package:image/image.dart' as image_lib;
+import 'package:tflite_flutter/tflite_flutter.dart';
+
+class IsolateInference {
+  static const String _debugName = "TFLITE_INFERENCE";
+  final ReceivePort _receivePort = ReceivePort();
+  late Isolate _isolate;
+  late SendPort _sendPort;
+
+  SendPort get sendPort => _sendPort;
+
+  Future<void> start() async {
+    _isolate = await Isolate.spawn<SendPort>(entryPoint, _receivePort.sendPort,
+        debugName: _debugName);
+    _sendPort = await _receivePort.first;
+  }
+
+  Future<void> close() async {
+    _isolate.kill();
+    _receivePort.close();
+  }
+
+  static void entryPoint(SendPort sendPort) async {
+    final port = ReceivePort();
+    sendPort.send(port.sendPort);
+
+    await for (final InferenceModel isolateModel in port) {
+      image_lib.Image? image;
+      image = isolateModel.image;
+      int inModelWidth = 640;
+      int inModelHeight = 640;
+      int numClasses = 6;
+
+      final imgResized = image_lib.copyResize(image!,
+          width: inModelWidth, height: inModelHeight);
+      final imgNormalized = List.generate(
+        inModelHeight,
+        (y) => List.generate(
+          inModelWidth,
+          (x) {
+            final pixel = imgResized.getPixel(x, y);
+            return [pixel.rNormalized, pixel.gNormalized, pixel.bNormalized];
+          },
+        ),
+      );
+
+      // output shape:
+      // 1 : batch size
+      // 4 + 6: left, top, right, bottom and probabilities for each class
+      // 8400: num predictions
+      final output = [
+        List<List<double>>.filled(4 + numClasses, List<double>.filled(8400, 0))
+      ];
+      int predictionTimeStart = DateTime.now().millisecondsSinceEpoch;
+      Interpreter interpreter =
+          Interpreter.fromAddress(isolateModel.interpreterAddress);
+
+      interpreter.run([imgNormalized], output);
+      debugPrint(
+          'Prediction time: ${DateTime.now().millisecondsSinceEpoch - predictionTimeStart} ms');
+
+      isolateModel.responsePort.send(output[0]);
+    }
+  }
+
+  static List<double> xywh2xyxy(List<double> bbox) {
+    double halfWidth = bbox[2] / 2;
+    double halfHeight = bbox[3] / 2;
+    return [
+      bbox[0] - halfWidth,
+      bbox[1] - halfHeight,
+      bbox[0] + halfWidth,
+      bbox[1] + halfHeight,
+    ];
+  }
+
+  /// Computes the intersection over union between two bounding boxes encoded with
+  /// the xyxy format.
+  static double computeIou(List<double> bbox1, List<double> bbox2) {
+    assert(bbox1[0] < bbox1[2]);
+    assert(bbox1[1] < bbox1[3]);
+    assert(bbox2[0] < bbox2[2]);
+    assert(bbox2[1] < bbox2[3]);
+
+    // Determine the coordinate of the intersection rectangle
+    double xLeft = max(bbox1[0], bbox2[0]);
+    double yTop = max(bbox1[1], bbox2[1]);
+    double xRight = min(bbox1[2], bbox2[2]);
+    double yBottom = min(bbox1[3], bbox2[3]);
+
+    if (xRight < xLeft || yBottom < yTop) {
+      return 0;
+    }
+    double intersectionArea = (xRight - xLeft) * (yBottom - yTop);
+    double bbox1Area = (bbox1[2] - bbox1[0]) * (bbox1[3] - bbox1[1]);
+    double bbox2Area = (bbox2[2] - bbox2[0]) * (bbox2[3] - bbox2[1]);
+
+    double iou = intersectionArea / (bbox1Area + bbox2Area - intersectionArea);
+    assert(iou >= 0 && iou <= 1);
+    return iou;
+  }
+}
+
+class InferenceModel {
+  image_lib.Image? image;
+  int interpreterAddress;
+  late SendPort responsePort;
+
+  InferenceModel(this.image, this.interpreterAddress);
+}

lib/utils/detection/nms.dart

@@ -74,6 +74,7 @@ import 'dart:math';
       sortedBestScores.removeAt(index);
     }
   }
+
   return (finalClasses, finalBboxes, finalScores);
 }
 

lib/utils/detection/yolo.dart

@@ -1,7 +1,10 @@
+import 'dart:isolate';
+
 import 'package:flutter/foundation.dart';
 import 'package:tflite_flutter/tflite_flutter.dart';
 import 'package:image/image.dart';
 
+import 'isolate_inference.dart';
 import 'nms.dart';
 
 class YoloModel {
@@ -10,6 +13,7 @@ class YoloModel {
   final int inHeight;
   final int numClasses;
   Interpreter? _interpreter;
+  late final IsolateInference isolateInference;
 
   YoloModel(
     this.modelPath,
@@ -20,11 +24,12 @@ class YoloModel {
 
   Future<void> init() async {
     _interpreter = await Interpreter.fromAsset(modelPath);
+    isolateInference = IsolateInference();
+    await isolateInference.start();
   }
 
   List<List<double>> infer(Image image) {
     assert(_interpreter != null, 'The model must be initialized');
-
     final imgResized = copyResize(image, width: inWidth, height: inHeight);
     final imgNormalized = List.generate(
       inHeight,
@@ -51,6 +56,21 @@ class YoloModel {
     return output[0];
   }
 
+  Future<List<List<double>>> _inference(InferenceModel inferenceModel) async {
+    ReceivePort responsePort = ReceivePort();
+    isolateInference.sendPort
+        .send(inferenceModel..responsePort = responsePort.sendPort);
+    // get inference result.
+    var results = await responsePort.first;
+    return results;
+  }
+
+  // inference still image
+  Future<List<List<double>>> inferenceImage(Image image) async {
+    var isolateModel = InferenceModel(image, _interpreter?.address ?? 0);
+    return _inference(isolateModel);
+  }
+
   Future<(List<int>, List<List<double>>, List<double>)> postprocess(
     List<List<double>> unfilteredBboxes,
     int imageWidth,