Add background removal pipeline

[xenova]

This PR makes it far easier to use background-removal models.

import { pipeline } from '@huggingface/transformers';

const segmenter = await pipeline('background-removal', 'onnx-community/BEN2-ONNX');
const url = 'https://huggingface.co/datasets/Xenova/transformers.js-docs/resolve/main/woman-with-afro_medium.jpg';
const output = await segmenter(url);
output[0].save('mask.png');

Input	Output

Tested with:

• ✅ https://huggingface.co/Xenova/modnet
• ✅ https://huggingface.co/briaai/RMBG-1.4
• ✅ https://huggingface.co/briaai/RMBG-2.0
• ✅ https://huggingface.co/onnx-community/mediapipe_selfie_segmentation
• ✅ https://huggingface.co/onnx-community/mediapipe_selfie_segmentation_landscape
• ✅ https://huggingface.co/onnx-community/modnet-webnn
• ✅ https://huggingface.co/onnx-community/BiRefNet-ONNX
• ✅ https://huggingface.co/onnx-community/BiRefNet_lite-ONNX
• ✅ https://huggingface.co/onnx-community/BiRefNet-portrait-ONNX
• ✅ https://huggingface.co/onnx-community/BiRefNet-COD-ONNX
• ✅ https://huggingface.co/onnx-community/BiRefNet-DIS5K-TR_TEs-ONNX
• ✅ https://huggingface.co/onnx-community/BiRefNet-DIS5K-ONNX
• ✅ https://huggingface.co/onnx-community/BiRefNet-HRSOD_DHU-ONNX
• ✅ https://huggingface.co/onnx-community/BiRefNet_512x512-ONNX
• ✅ https://huggingface.co/onnx-community/ormbg-ONNX
• ✅ https://huggingface.co/onnx-community/BEN2-ONNX
• ✅ https://huggingface.co/onnx-community/MVANet-ONNX
• ✅ https://huggingface.co/onnx-community/ISNet-ONNX

TODO:

• Update model cards

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[JonasDoesThings]

Thanks so much for the work @xenova!

Could you maybe share your testing code that worked for briaai/RMBG-2.0? Trying to get it to work myself, but creating the pipeline fails with a strange numeric error and the PR only contains tests for 2 other models 😅

[TrumanDu]

@JonasDoesThings u can try edit model config.json:

{
  "_name_or_path": "ZhengPeng7/BiRefNet",
  "architectures": [
    "BiRefNet"
  ],
  "auto_map": {
    "AutoConfig": "BiRefNet_config.BiRefNetConfig",
    "AutoModelForImageSegmentation": "birefnet.BiRefNet"
  },
  "custom_pipelines": {
    "image-segmentation": {
      "pt": [
        "AutoModelForImageSegmentation"
      ],
      "tf": [],
      "type": "image"
    }
  },
  "bb_pretrained": false,
  "model_type": "birefnet"
}

[ZhangPeng4242]

@JonasDoesThings u can try edit model config.json:

{
  "_name_or_path": "ZhengPeng7/BiRefNet",
  "architectures": [
    "BiRefNet"
  ],
  "auto_map": {
    "AutoConfig": "BiRefNet_config.BiRefNetConfig",
    "AutoModelForImageSegmentation": "birefnet.BiRefNet"
  },
  "custom_pipelines": {
    "image-segmentation": {
      "pt": [
        "AutoModelForImageSegmentation"
      ],
      "tf": [],
      "type": "image"
    }
  },
  "bb_pretrained": false,
  "model_type": "birefnet"
}

did you make it work? I tried but still not working.

[TrumanDu]

@JonasDoesThings u can try edit model config.json:

{
  "_name_or_path": "ZhengPeng7/BiRefNet",
  "architectures": [
    "BiRefNet"
  ],
  "auto_map": {
    "AutoConfig": "BiRefNet_config.BiRefNetConfig",
    "AutoModelForImageSegmentation": "birefnet.BiRefNet"
  },
  "custom_pipelines": {
    "image-segmentation": {
      "pt": [
        "AutoModelForImageSegmentation"
      ],
      "tf": [],
      "type": "image"
    }
  },
  "bb_pretrained": false,
  "model_type": "birefnet"
}

did you make it work? I tried but still not working.

yeah, I'm ok.

[sliterok]

For anyone else stumbling here from the #1107
To make it work you have to add
"model_type": "birefnet"
at the end of the node_modules/.pnpm/@huggingface+transformers@3.5.1/node_modules/@huggingface/transformers/.cache/briaai/RMBG-2.0/config.json