Credits and third-party notices
UNBG's own code is proprietary. It relies on the following work, used under the licences shown.
Background-removal model
BiRefNet lite 1024 (fp16, WebGPU): jiabins0303/birefnet-lite-1024-webgpu, based on ZhengPeng7/BiRefNet_lite. Licence declared by the publisher: MIT. BiRefNet by Peng Zheng et al. (MIT, Copyright (c) 2024 ZhengPeng); ONNX conversion by jiabins0303.
BiRefNet lite 512 (fp16): studioludens/birefnet-lite-512, based on ZhengPeng7/BiRefNet_lite. Licence declared by the publisher: MIT. BiRefNet by Peng Zheng et al. (MIT, Copyright (c) 2024 ZhengPeng); ONNX conversion by studioludens.
U²-Netp 320 (fp32, phones and tablets): danielgatis/rembg, based on xuebinqin/U-2-Net. Licence declared by the publisher: Apache-2.0. U²-Net by Xuebin Qin et al. (Apache-2.0); U²-Netp ONNX export distributed by the rembg project (danielgatis/rembg).
The training data of BiRefNet_lite, the model used on computers, includes the DIS5K dataset. The dataset's terms of use say that without the original authors' permission, "commercial use of this dataset is prohibited even after copying, editing, processing or any operations of this database". Whether that applies to the trained model is unresolved, so commercial use of results hasn't been confirmed during the beta.
Software
UNBG runs the model with ONNX Runtime Web 1.30.0 and uses other open-source libraries. Their copyright and licence notices are reproduced in full in third-party-notices.txt.
Homepage example photograph
“Golden Retriever Carlos” by Dirk Vorderstraße, via Wikimedia Commons, licensed under CC BY 2.0. Resized, and the background removed with UNBG. The photographer does not endorse UNBG.
.jpg){kind=link}