Image Segmentation
Transformers.js
ONNX
BiRefNet
swin
background-removal
dichotomous-image-segmentation
webgpu
onnxruntime-web
Instructions to use CardAnvilModels/BiRefNet-lite-1024-WebGPU with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers.js
How to use CardAnvilModels/BiRefNet-lite-1024-WebGPU with Transformers.js:
// npm i @huggingface/transformers import { pipeline } from '@huggingface/transformers'; // Allocate pipeline const pipe = await pipeline('image-segmentation', 'CardAnvilModels/BiRefNet-lite-1024-WebGPU'); - BiRefNet
How to use CardAnvilModels/BiRefNet-lite-1024-WebGPU with BiRefNet:
# Option 1: use with transformers from transformers import AutoModelForImageSegmentation birefnet = AutoModelForImageSegmentation.from_pretrained("CardAnvilModels/BiRefNet-lite-1024-WebGPU", trust_remote_code=True)# Option 2: use with BiRefNet # Install from https://github.com/ZhengPeng7/BiRefNet from models.birefnet import BiRefNet model = BiRefNet.from_pretrained("CardAnvilModels/BiRefNet-lite-1024-WebGPU") - Notebooks
- Google Colab
- Kaggle
File size: 391 Bytes
90abcec | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 | {
"do_normalize": true,
"do_rescale": true,
"do_resize": true,
"feature_extractor_type": "ViTFeatureExtractor",
"image_mean": [
0.485,
0.456,
0.406
],
"image_processor_type": "ViTFeatureExtractor",
"image_std": [
0.229,
0.224,
0.225
],
"resample": 2,
"rescale_factor": 0.00392156862745098,
"size": {
"height": 1024,
"width": 1024
}
} |