hyuto / yolov8-seg-tfjs

YOLOv8 Segmentation Right in The Browser Using Tensorflow.js

Object Segmentation Application Using YOLOv8 and Tensorflow.js

---

⚠️ Low performance on doing video processing, need to be improved.

Object Segmentation application right in your browser. Serving YOLOv8-Seg in browser using tensorflow.js with webgl backend.

TO DO

1. Readable code
2. Improve performance

Setup

git clone https://github.com/Hyuto/yolov8-seg-tfjs.git
cd yolov8-seg-tfjs
yarn install #Install dependencies

Scripts

yarn start # Start dev server
yarn build # Build for productions

Model

YOLOv8n-Seg model converted to tensorflow.js.

used model : yolov8n-seg
size       : 14 Mb

Reference

• https://github.com/ultralytics/ultralytics
• https://github.com/Hyuto/yolov8-seg-onnxruntime-web

yolov8-seg-tfjs's Issues

How to obtain image data based on mask and crop each object

How to obtain image data based on mask and crop each object? Thank you very much

Failed to run when it is a single class

Hi Hyuto,

Thanks for the awesome job. I would like to highlight an issue I faced

const [scores, classes] = tf.tidy(() => {
    const rawScores = transRes.slice([0, 4], [-1, numClass]).squeeze(); // [n, 1]
    return [rawScores.max(1), rawScores.argMax(1)];
  }); // get scores and classes

when there is only 1 class, the above code will fail.

Would really appreciate if you can help fix it.

Thanks

yolov8x-seg work error

yolov8x-seg work error

how fix
thanks

Hello, I want to know how I can get the tracking id

Hello, I want to know how I can get the tracking id, I don't contact yolo much, I read your source code, may I ask where is the document of his output format

Is it possible to load a custom model?