ISNet General ONNX — local recovery model

Publisher: IMG.LY
Source: https://huggingface.co/imgly/isnet-general-onnx
Revision: 440dea96dd4a3b06bbbf5abec3e26569dd7ec49f
Artifact: onnx/model_fp16.onnx (88,152,708 bytes)
SHA-256: 2eb4b5dda7ec41c617e59706e5aafa1f978c9a5f983d2518d9f0ae4d6eb04f20

The publisher's model repository declares MIT in its license metadata.
That repository does not include a substantive license text or provenance
statement. Public-release rights/notice review remains outstanding; this
record does not assert that the publisher's metadata alone completes it.

Upstream research: Highly Accurate Dichotomous Image Segmentation (ECCV 2022)
Xuebin Qin, Hang Dai, Xiaobin Hu, Deng-Ping Fan, Ling Shao, Luc Van Gool.
https://github.com/xuebinqin/DIS
The upstream repository identifies its code and evaluation metric as
Apache-2.0. Reqly does not incorporate that Python code or the IMG.LY
background-removal wrapper. The TypeScript adapter uses ONNX Runtime.