Instructions to use Swedemom/ben2-fp32-onnx with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers.js
How to use Swedemom/ben2-fp32-onnx with Transformers.js:
// npm i @huggingface/transformers import { pipeline } from '@huggingface/transformers'; // Allocate pipeline const pipe = await pipeline('image-segmentation', 'Swedemom/ben2-fp32-onnx');
BEN2 (full precision) for Transformers.js
BEN2 by Prama LLC, MIT licensed, as a single full-precision (fp32) ONNX file for Transformers.js in the browser.
It is the half-precision onnx-community/BEN2-ONNX model converted to fp32. The half-precision file fails on Apple graphics cards under WebGPU (its LayerNorm shader won't compile); this one runs on Mac and Windows.
Used by the Swedemom Hub photo editor's Remove Background tool. The model runs in the imager's browser; no photo leaves the station.
onnx/model.onnx: 422,453,328 bytes, SHA-256e18a47e283826cc5f9a90c7ae6ca684fe8e633b3f07bd3e550cc8b4be7bb99aa- Input: 1 x 3 x 1024 x 1024, ImageNet mean and std. Output: the foreground mask.
License: MIT, see LICENSE (Copyright (c) 2025 Prama LLC).
- Downloads last month
- -
Model tree for Swedemom/ben2-fp32-onnx
Base model
PramaLLC/BEN2