skillsafe-ai/u2net

U^2-Net salient object / background removal

Browser-ready import artifacts for background-removal, produced by SkillSafe's reproducible converter (models/ in skillsafe.ai) from a pinned upstream source. Every byte here is derivable from that source plus the recipe below; nothing was edited by hand.

Provenance

Upstream	https://github.com/danielgatis/rembg/releases/download/v0.0.0/u2net.onnx
Upstream SHA-256 / commit	v0.0.0
Recipe	recipes/u2net.yaml — sha256 7e8f815db7a9b8797008d8de0d9ef3bc1a1976d5d06914cea8430ae1d0e10441
Toolchain	Python 3.12.13, torch 2.10.0, onnx 1.23.0, onnxruntime 1.30.0 on Darwin 25.6.0 arm64
Converted	2026-09-22T18:56:11+00:00

Files

file	class	size	SHA-256
u2net.onnx	registry	167.84 MB	8d10d2f3bb75ae3b6d527c77944fc5e7dcd94b29809d47a739a7a728a912b491

registry files are parameter files served from models.skillsafe.ai once vetted; bundle files ship inside an app; registry-shared is a runtime library reused by every model of the same architecture.

Verification

Imported as published upstream (no conversion). Each file is pinned by SHA-256 to its source; every ONNX file passed onnx.checker and a CPU smoke run under onnxruntime with zero-filled inputs at the declared shapes:

file	inputs	outputs	ms
u2net.onnx	input.1[1, 3, 320, 320]	1959[1, 1, 320, 320], 1960[1, 1, 320, 320], 1961[1, 1, 320, 320], 1962[1, 1, 320, 320], 1963[1, 1, 320, 320], 1964[1, 1, 320, 320], 1965[1, 1, 320, 320]	1050.6

Use in the browser

import * as ort from "onnxruntime-web";
const session = await ort.InferenceSession.create("https://huggingface.co/skillsafe-ai/u2net/resolve/main/u2net.onnx", { executionProviders: ["webgpu", "wasm"] });

Contract (u2net.onnx): input input.1 float32 [1, 3, 320, 320] → output 1959 float32 [1, 1, 320, 320], 1960 float32 [1, 1, 320, 320], 1961 float32 ['Sigmoid1961_dim_0', 'Sigmoid1961_dim_1', 'Sigmoid1961_dim_2', 'Sigmoid1961_dim_3'], 1962 float32 ['Sigmoid1962_dim_0', 'Sigmoid1962_dim_1', 'Sigmoid1962_dim_2', 'Sigmoid1962_dim_3'], 1963 float32 ['Sigmoid1963_dim_0', 'Sigmoid1963_dim_1', 'Sigmoid1963_dim_2', 'Sigmoid1963_dim_3'], 1964 float32 ['Sigmoid1964_dim_0', 'Sigmoid1964_dim_1', 'Sigmoid1964_dim_2', 'Sigmoid1964_dim_3'], 1965 float32 ['Sigmoid1965_dim_0', 'Sigmoid1965_dim_1', 'Sigmoid1965_dim_2', 'Sigmoid1965_dim_3']. Opset 11.

Licence and attribution

U^2-Net: Xuebin Qin et al., Apache License 2.0 (https://github.com/xuebinqin/U-2-Net). ONNX distributed by rembg (Daniel Gatis, MIT). https://github.com/danielgatis/rembg

Licence: Apache-2.0 — notice: https://github.com/xuebinqin/U-2-Net/blob/master/LICENSE. The conversion recipe and this model card are part of the SkillSafe repository and carry its licence; the weights remain under the upstream licence above.

The full manifest.json in this repo records the recipe, sources, toolchain (including the uv.lock hash) and per-file verification numbers.