Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/shaqian/tfjs-yolo
YOLO v3 and Tiny YOLO v1, v2, v3 with Tensorflow.js
https://github.com/shaqian/tfjs-yolo
tiny-yolo yolo yolo-tiny yolov2 yolov3
Last synced: about 2 months ago
JSON representation
YOLO v3 and Tiny YOLO v1, v2, v3 with Tensorflow.js
- Host: GitHub
- URL: https://github.com/shaqian/tfjs-yolo
- Owner: shaqian
- License: mit
- Created: 2018-10-28T14:31:31.000Z (about 6 years ago)
- Default Branch: master
- Last Pushed: 2019-04-11T11:40:51.000Z (over 5 years ago)
- Last Synced: 2024-10-29T22:54:11.822Z (about 2 months ago)
- Topics: tiny-yolo, yolo, yolo-tiny, yolov2, yolov3
- Language: JavaScript
- Homepage: https://www.npmjs.com/package/tfjs-yolo
- Size: 77.1 KB
- Stars: 111
- Watchers: 8
- Forks: 19
- Open Issues: 3
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
- awesome-yolo-object-detection - shaqian/tfjs-yolo - yolo?style=social"/> : YOLO v3 and Tiny YOLO v1, v2, v3 with Tensorflow.js. (Other Versions of YOLO)
- awesome-yolo-object-detection - shaqian/tfjs-yolo - yolo?style=social"/> : YOLO v3 and Tiny YOLO v1, v2, v3 with Tensorflow.js. (Other Versions of YOLO)
README
# tfjs-yolo
[`YOLO`](https://pjreddie.com/darknet/yolo/) object detection with [Tensorflow.js](https://js.tensorflow.org/). Supports YOLO v3 and Tiny YOLO v1, v2, v3.
## Demo
- Detect objects using your webcam:
https://shaqian.github.io/tfjs-yolo-demo/
- Not hotdog PWA: https://shaqian.github.io/Not-Hotdog/
## Install
```
npm install tfjs-yolo
```
## Usage
### Import module
```javascript
import yolo from 'tfjs-yolo';
```
### Initialize and load model
```javascript
// Use default models (stored in my GitHub demo repo)
let myYolo = await yolo.v1tiny();
let myYolo = await yolo.v2tiny();
let myYolo = await yolo.v3tiny();
let myYolo = await yolo.v3();
// or specify path or handler, see https://js.tensorflow.org/api/0.13.3/#loadModel
let myYolo = await yolo.v3tiny("https://.../model.json");
// or use frozen model, see https://js.tensorflow.org/api/0.13.3/#loadFrozenModel
let myYolo = await yolo.v3tiny(
"https://.../weights_manifest.json",
"https://.../tensorflowjs_model.pb"
);
```
### Run model
Supported input html element:
- img
- canvas
- video
```javascript
const boxes = await myYolo.predict(canvas);
// Optional settings
const boxes = await myYolo.predict(
canvas,
{
maxBoxes: 5, // defaults to 20
scoreThreshold: .2, // defaults to .5
iouThreshold: .5, // defaults to .3
numClasses: 80, // defaults to 80 for yolo v3, tiny yolo v2, v3 and 20 for tiny yolo v1
anchors: [...], // See ./src/config.js for examples
classNames: [...], // defaults to coco classes for yolo v3, tiny yolo v2, v3 and voc classes for tiny yolo v1
inputSize: 416, // defaults to 416
}
);
```
### Output box format
```javascript
{
top, // Float
left, // Float
bottom, // Float
right, // Float
height, // Float
width, // Float
score, // Float
class // String, e.g. person
}
```
### Dispose model
```javascript
myYolo.dispose();
```
## Credits
- https://github.com/qqwweee/keras-yolo3
- https://github.com/zqingr/tfjs-yolov3
- https://github.com/ModelDepot/tfjs-yolo-tiny
- https://github.com/allanzelener/YAD2K