pytorch_lite 3.0.0

Flutter package to help run pytorch lite models classification and yolov5 and yolov8

pytorch_lite

• flutter package to help run pytorch lite models classification and yolov5
• ios support (can be added following this https://github.com/pytorch/ios-demo-app) PR will be appreciated

example for Classification

example for Object detection

Usage

preparing the model

• classification

import torch
from torch.utils.mobile_optimizer import optimize_for_mobile


model = torch.load('model_scripted.pt',map_location="cpu")
model.eval()
example = torch.rand(1, 3, 224, 224)
traced_script_module = torch.jit.trace(model, example)
optimized_traced_model = optimize_for_mobile(traced_script_module)
optimized_traced_model._save_for_lite_interpreter("model.pt")

• 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

• object detection (yolov8)

!yolo mode=export model="your model" format=torchscript optimize

example

!yolo mode=export model=yolov8s.pt format=torchscript 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_classification.pt
 - assets/labels_classification.txt
 - 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:pytorch_lite/pytorch_lite.dart';

Load model

Either classification model:

ClassificationModel classificationModel= await PytorchLite.loadClassificationModel(
          "assets/models/model_classification.pt", 224, 224,
          labelPath: "assets/labels/label_classification_imageNet.txt");

Or objectDetection model:

ModelObjectDetection objectModel = await PytorchLite.loadObjectDetectionModel(
          "assets/models/yolov5s.torchscript", 80, 640, 640,
          labelPath: "assets/labels/labels_objectDetection_Coco.txt",
          objectDetectionModelType: ObjectDetectionModelType.yolov5);

Get classification prediction as label

String imagePrediction = await classificationModel.getImagePrediction(await File(image.path).readAsBytes());

Get classification prediction as raw output layer

List<double?>? predictionList = await _imageModel!.getImagePredictionList(
      await File(image.path).readAsBytes(),
    );

Get classification prediction as Probabilities (incase model is not using softmax)

List<double?>? predictionListProbabilites = await _imageModel!.getImagePredictionListProbabilities(
      await File(image.path).readAsBytes(),
    );

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)

Image prediction for an image with custom mean and std

final mean = [0.5, 0.5, 0.5];
final std = [0.5, 0.5, 0.5];
String prediction = await classificationModel
        .getImagePrediction(image, mean: mean, std: std);

#References

• Code used the same strucute as the package https://pub.dev/packages/pytorch_mobile
• While using the updated code from https://github.com/pytorch/android-demo-app