Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -4,79 +4,4 @@ library_name: transformers.js
|
|
| 4 |
|
| 5 |
https://huggingface.co/nielsr/slimsam-77-uniform with ONNX weights to be compatible with Transformers.js.
|
| 6 |
|
| 7 |
-
## Usage (Transformers.js)
|
| 8 |
-
|
| 9 |
-
If you haven't already, you can install the [Transformers.js](https://huggingface.co/docs/transformers.js) JavaScript library from [NPM](https://www.npmjs.com/package/@xenova/transformers) using:
|
| 10 |
-
```bash
|
| 11 |
-
npm i @xenova/transformers
|
| 12 |
-
```
|
| 13 |
-
|
| 14 |
-
**Example:** Perform mask generation with `Xenova/slimsam-77-uniform`.
|
| 15 |
-
|
| 16 |
-
```js
|
| 17 |
-
import { SamModel, AutoProcessor, RawImage } from '@xenova/transformers';
|
| 18 |
-
|
| 19 |
-
// Load model and processor
|
| 20 |
-
const model = await SamModel.from_pretrained('Xenova/slimsam-77-uniform');
|
| 21 |
-
const processor = await AutoProcessor.from_pretrained('Xenova/slimsam-77-uniform');
|
| 22 |
-
|
| 23 |
-
// Prepare image and input points
|
| 24 |
-
const img_url = 'https://huggingface.co/datasets/Xenova/transformers.js-docs/resolve/main/corgi.jpg';
|
| 25 |
-
const raw_image = await RawImage.read(img_url);
|
| 26 |
-
const input_points = [[[340, 250]]] // 2D localization of a window
|
| 27 |
-
|
| 28 |
-
// Process inputs and perform mask generation
|
| 29 |
-
const inputs = await processor(raw_image, input_points);
|
| 30 |
-
const outputs = await model(inputs);
|
| 31 |
-
|
| 32 |
-
// Post-process masks
|
| 33 |
-
const masks = await processor.post_process_masks(outputs.pred_masks, inputs.original_sizes, inputs.reshaped_input_sizes);
|
| 34 |
-
console.log(masks);
|
| 35 |
-
// [
|
| 36 |
-
// Tensor {
|
| 37 |
-
// dims: [ 1, 3, 410, 614 ],
|
| 38 |
-
// type: 'bool',
|
| 39 |
-
// data: Uint8Array(755220) [ ... ],
|
| 40 |
-
// size: 755220
|
| 41 |
-
// }
|
| 42 |
-
// ]
|
| 43 |
-
const scores = outputs.iou_scores;
|
| 44 |
-
console.log(scores);
|
| 45 |
-
// Tensor {
|
| 46 |
-
// dims: [ 1, 1, 3 ],
|
| 47 |
-
// type: 'float32',
|
| 48 |
-
// data: Float32Array(3) [
|
| 49 |
-
// 0.8350210189819336,
|
| 50 |
-
// 0.9786665439605713,
|
| 51 |
-
// 0.8379436731338501
|
| 52 |
-
// ],
|
| 53 |
-
// size: 3
|
| 54 |
-
// }
|
| 55 |
-
```
|
| 56 |
-
|
| 57 |
-
You can then visualize the generated mask with:
|
| 58 |
-
|
| 59 |
-
```js
|
| 60 |
-
const image = RawImage.fromTensor(masks[0][0].mul(255));
|
| 61 |
-
image.save('mask.png');
|
| 62 |
-
```
|
| 63 |
-
|
| 64 |
-
|
| 65 |
-
|
| 66 |
-
Next, select the channel with the highest IoU score, which in this case is the second (green) channel. Intersecting this with the original image gives us an isolated version of the subject:
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
## Demo
|
| 73 |
-
|
| 74 |
-
We've also got an online demo, which you can try out [here](https://huggingface.co/spaces/Xenova/segment-anything-web).
|
| 75 |
-
|
| 76 |
-
|
| 77 |
-
<video controls autoplay src="https://cdn-uploads.huggingface.co/production/uploads/61b253b7ac5ecaae3d1efe0c/Y0wAOw6hz9rWpwiuMoz2A.mp4"></video>
|
| 78 |
-
|
| 79 |
-
---
|
| 80 |
-
|
| 81 |
-
|
| 82 |
Note: Having a separate repo for ONNX weights is intended to be a temporary solution until WebML gains more traction. If you would like to make your models web-ready, we recommend converting to ONNX using [🤗 Optimum](https://huggingface.co/docs/optimum/index) and structuring your repo like this one (with ONNX weights located in a subfolder named `onnx`).
|
|
|
|
| 4 |
|
| 5 |
https://huggingface.co/nielsr/slimsam-77-uniform with ONNX weights to be compatible with Transformers.js.
|
| 6 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 7 |
Note: Having a separate repo for ONNX weights is intended to be a temporary solution until WebML gains more traction. If you would like to make your models web-ready, we recommend converting to ONNX using [🤗 Optimum](https://huggingface.co/docs/optimum/index) and structuring your repo like this one (with ONNX weights located in a subfolder named `onnx`).
|