Cocoa detector and classification with Yolov5 and flutter

Send message if you want the full code

Testing

Results

  1. Confusion matrix

2. Precision Confidence Curve

  1. Others

Train

Val

flutter_pytorch

  • flutter plugin to help run pytorch lite models classification for example yolov5 doesn’t give support for yolov7
  • ios support (can be added following this https://github.com/pytorch/ios-demo-app) PR will be appreciated

Getting Started

Usage

preparing the model

  • object detection (yolov5)
!python export.py --weights "the weights of your model" --include torchscript --img 640 --optimize

example

!python export.py --weights yolov5s.pt --include torchscript --img 640 --optimize

Installation

To use this plugin, add pytorch_lite as a dependency in your pubspec.yaml file. Create a assets folder with your pytorch model and labels if needed. Modify pubspec.yaml accordingly.

assets:
 - assets/models/model_objectDetection.torchscript
 - assets/labels_objectDetection.txt

Run flutter pub get

For release

  • Go to android/app/build.gradle
  • Add those next lines in the release config

shrinkResources false
minifyEnabled false

example

    buildTypes {
        release {
            shrinkResources false
            minifyEnabled false
            // TODO: Add your own signing config for the release build.
            // Signing with the debug keys for now, so `flutter run --release` works.
            signingConfig signingConfigs.debug
        }
    }

Import the library

import 'package:flutter_pytorch/flutter_pytorch.dart';

Load model

Either classification model:

ObjectDetection model:
```dart
ModelObjectDetection objectModel = await FlutterPytorch.loadObjectDetectionModel(
          "assets/models/yolov5s.torchscript", 80, 640, 640,
          labelPath: "assets/labels/labels_objectDetection_Coco.txt");

Get object detection prediction for an image

 List<ResultObjectDetection?> objDetect = await _objectModel.getImagePrediction(await File(image.path).readAsBytes(),
        minimumScore: 0.1, IOUThershold: 0.3);

Get render boxes with image

objectModel.renderBoxesOnImage(_image!, objDetect)

#References

Get render boxes with image

GitHub

View Github