NORI Tools — Image background removal

U²-Net / U2NETP
Authors: Xuebin Qin, Zichen Zhang, Chenyang Huang, Masood Dehghan,
Osmar R. Zaiane and Martin Jagersand.
Source: https://github.com/xuebinqin/U-2-Net
License: Apache License 2.0 (full text: licenses/U-2-Net.txt)
Unmodified ONNX conversion distributed by rembg:
https://github.com/danielgatis/rembg/releases/download/v0.0.0/u2netp.onnx
SHA-256: 309c8469258dda742793dce0ebea8e6dd393174f89934733ecc8b14c76f4ddd8
Model architecture and pretrained model are provided by the U²-Net project.
The conversion is used without changing the model weights.

ONNX Runtime Web 1.20.1
Copyright (c) Microsoft Corporation. All rights reserved.
Source: https://github.com/microsoft/onnxruntime/tree/v1.20.1
License: MIT (full text: licenses/ONNX-Runtime.txt)
Runtime JavaScript and WebAssembly are loaded from jsDelivr.

Both licenses permit commercial use subject to their terms, including
retention of required license and attribution notices. No paid inference
service or API key is used. No image data is sent to a server by this tool.
The U2NETP model is served by this site and cached by the browser when possible.

Standard U2NET model (quality mode), also from the U²-Net project:
https://github.com/danielgatis/rembg/releases/download/v0.0.0/u2net.onnx
Pinned mirror: https://huggingface.co/Heliosoph/u2net-onnx/tree/7fc34deee10329bc039c10a73b98090d0c6f5c59
SHA-256: 8d10d2f3bb75ae3b6d527c77944fc5e7dcd94b29809d47a739a7a728a912b491
This model (approximately 176 MB) is downloaded directly to the browser
from Hugging Face. No user image is uploaded. The mirror file is checked
against the SHA-256 of the upstream ONNX weights before first use.