#636 · Primary category: Computer Vision

tfjs-yolo-tiny

browser computer-vision deep-learning detection machine-learning npm-package object-detection tensorflow tensorflow-js tfjs yolo yolo-models

In-Browser Object Detection using Tiny YOLO on Tensorflow.js

Project last updated:03/25/19

GitHub Stars

533

Forks

88

Contributors

2

License

MIT

Why we included this project

Object detection usually means shipping frames to a server, but this package keeps the whole pipeline in the browser. It wraps Tiny YOLO for TensorFlow.js with a pre-trained model downloader and a small API: feed it a 416x416 image tensor and you get back bounding boxes with class names and confidence scores, which is enough to hook up webcam detection in a few lines of JavaScript. The thresholds for class probability, IoU, and box filtering are all configurable, so you can tighten or loosen detection to fit your scene. It runs at roughly a frame per second on a mid-2014 MacBook Pro, so it suits demos, on-device privacy-sensitive tools, and learning how YOLO inference works rather than production video analysis. Note that it targets the older tfjs API and hasn't been updated since 2019.

Articles for this project

No articles for this project yet.

To suggest a topic or contribute an article, contact us.

Related projects in this category