|
Download README.md from jitendra25/maya-models: direct link, hf CLI and curl.
- Browser
- Download file 1.66 kB
-
https://huggingface.co/jitendra25/maya-models/resolve/main/README.md
- Command line
-
hf download hf://jitendra25/maya-models/README.md
-
curl -L -o README.md https://huggingface.co/jitendra25/maya-models/resolve/main/README.md
1.66 kB
| license: other | |
| license_name: mixed-see-model-card | |
| tags: | |
| - onnx | |
| - face-swap | |
| # Maya models | |
| The model files that [Maya](https://github.com/jitendra205saini/Maya) downloads on first run. They are collected here so that the app's setup has one place to fetch them from. None of them was trained by this project. | |
| | File | What it does | Origin | | |
| |---|---|---| | |
| | `det_10g.onnx` | Face detection (RetinaFace) | InsightFace | | |
| | `w600k_r50.onnx` | Face recognition (ArcFace) | InsightFace | | |
| | `inswapper_128.fp16.onnx` | Face swapping | InsightFace | | |
| | `2d106det.onnx` | Face landmarks | InsightFace | | |
| | `scrfd_2.5g_bnkps.onnx` | Face detection (SCRFD) | InsightFace | | |
| | `yoloface_8n.onnx` | Face detection (YOLO) | YOLOv8-face | | |
| | `res50.onnx` | Face detection (RetinaFace ResNet50) | Pytorch_Retinaface | | |
| | `GFPGANv1.4.onnx` | Face restoration | GFPGAN (TencentARC) | | |
| | `GPEN-BFR-256.onnx`, `GPEN-BFR-512.onnx` | Face restoration | GPEN | | |
| | `codeformer_fp16.onnx` | Face restoration | CodeFormer | | |
| | `occluder.onnx` | Occlusion mask | Rope | | |
| | `faceparser_fp16.onnx` | Face parsing mask | face-parsing.PyTorch | | |
| | `rd64-uni-refined.pth` | Text-prompted mask (CLIPSeg) | CLIPSeg | | |
| ## Licences | |
| Each file keeps the licence of its original authors. Several of them, including the InsightFace models (`inswapper_128`, `det_10g`, `w600k_r50`, `2d106det`, `scrfd_2.5g_bnkps`) and CodeFormer, are released for non-commercial research use only. Check the licence of every model you use before using it for anything else. | |
| ## Responsible use | |
| Only swap faces of people who have agreed to it. Do not use these models to deceive, harass or impersonate anyone. | |