Credits
Open-source credits
Backgrounds Remover is built on open research and open-source software. These are the components that run in your browser, with their licenses.
IS-Net (DIS: Highly Accurate Dichotomous Image Segmentation)
Copyright Xuebin Qin and co-authors. Source: github.com/xuebinqin/DIS. The model architecture code is licensed under the Apache License 2.0. We export the network to ONNX ourselves from the upstream architecture and the authors' published general-use weights.
Xuebin Qin, Hang Dai, Xiaobin Hu, Deng-Ping Fan, Ling Shao, Luc Van Gool. "Highly Accurate Dichotomous Image Segmentation." ECCV 2022.
ONNX Runtime
Copyright Microsoft Corporation. Source: github.com/microsoft/onnxruntime, version 1.30.0, compiled from source to WebAssembly. Licensed under the MIT License. ONNX Runtime statically includes further open-source components; their notices are in the ONNX Runtime third-party notices.
Sample photo
The sample photo on the home page is "coffee" by Rachel Michetti from the scikit-image sample data, dedicated to the public domain under CC0.
Emscripten
Copyright the Emscripten authors. Source: github.com/emscripten-core/emscripten, version 4.0.23. The compiled C library and JavaScript glue in the engine are covered by the MIT and University of Illinois/NCSA licenses.
Dawn and emdawnwebgpu (WebGPU builds only)
Copyright the Dawn and Tint authors. Source: github.com/google/dawn. Licensed under the BSD 3-Clause License; the WebGPU JavaScript bindings carry the Emscripten MIT and NCSA licenses.