Licenses
| Model | License | Size |
|---|---|---|
| BiRefNet Lite | MIT | 109 MB |
| U²-Netp | Apache-2.0 | 4.4 MB |
| SlimSAM-77 | Apache-2.0 | 13 MB |
| Real-ESRGAN General x4v3 | BSD-3-Clause | 4.6 MB |
| MI-GAN (Places2, 512) | MIT | 27 MB |
| RF-DETR Nano (COCO, q8) | Apache-2.0 | 27 MB |
| PP-OCRv6 tiny (detección de texto) | Apache-2.0 | 1.7 MB |
| Depth Anything V2 Small (q8) | Apache-2.0 | 26 MB |
| NAFNet REDS-width64 (int8) | MIT | 87 MB |
- BiRefNet © Peng Zheng et al. — MIT. ONNX export by onnx-community.
- U²-Net / U²-Netp © Xuebin Qin et al. — Apache-2.0. ONNX file distributed by the rembg project (MIT).
- SlimSAM (compressed Segment Anything) © Zigeng Chen et al., based on Meta’s SAM — Apache-2.0. ONNX by Xenova/slimsam-77-uniform.
- Real-ESRGAN © Xintao Wang et al. — BSD-3-Clause. Our own ONNX conversion from the official weights.
- MI-GAN © 2024 Picsart AI Research — MIT (code and weights). ONNX from the andraniksargsyan/migan repository.
- RF-DETR © Roboflow — Apache-2.0. Quantized ONNX from onnx-community/rfdetr_nano-ONNX.
- PP-OCRv6 (PaddleOCR) © PaddlePaddle Authors — Apache-2.0. Official ONNX by PaddlePaddle.
- Depth Anything V2 Small © Lihe Yang et al. — Apache-2.0 (Small version only). Quantized ONNX by onnx-community.
- NAFNet © 2022 megvii-model (Megvii Research) — MIT. ONNX with int8 weights by OpenCV (opencv/deblurring_nafnet, MIT).
- PixelOE © Shin-Ying Yeh (KohakuBlueleaf) — Apache-2.0. Non-AI algorithm, ported to TypeScript for Pixelate (pixel art).
- ONNX Runtime Web © Microsoft — MIT.
- Vue — MIT. Lucide icons — ISC. Figtree font © Erik Kennedy — SIL Open Font License 1.1.