face-analyzer-weights
Mirror of the permissively licensed model files used by
multimodel-face-analyzer.
The app downloads them on demand with tools/fetch_models.py, which checks each file's sha256
against models/manifest.json in that repository.
Only files whose code and weights both clearly allow redistribution (MIT, Apache-2.0, BSD-2-Clause, CC BY 4.0, or the BSD-style Intel license) are hosted here. Restricted or unclearly licensed weights are not mirrored: the fetch script downloads them from their original upstream source, or tells you where to get them yourself.
Each file keeps the license of its upstream project, listed below. Nothing here changes those terms, and no file in this repository is relicensed. Files are redistributed as published upstream (some renamed); where a file is a conversion, its attribution line says so.
Files
| File | Bytes | Used for | Upstream source | License |
|---|---|---|---|---|
opencv_face_detector.pbtxt |
34,911 | face detection | github.com/opencv/opencv | Apache-2.0 |
retinaface_mobilenet0.25.onnx |
1,767,506 | face detection | huggingface.co/amd/retinaface | MIT (biubug6/Pytorch_Retinaface); ONNX re-export Apache-2.0 (amd/retinaface) |
fairface_7class.onnx |
85,160,429 | age, gender, race | github.com/dchen236/FairFace | CC BY 4.0 |
mivolo_v2.safetensors |
115,078,528 | age, gender | huggingface.co/iitolstykh/mivolo_v2 | Apache-2.0 |
mivolo_v2_config.json |
225 | age, gender | huggingface.co/iitolstykh/mivolo_v2 | Apache-2.0 |
emotion_ferplus.onnx |
35,040,571 | emotion | github.com/onnx/models | MIT |
mini_xception_fer.h5 |
872,856 | emotion | github.com/oarriaga/face_classification | MIT |
face_landmarker.task |
3,758,596 | face landmarks, gaze, liveness | ai.google.dev/edge/mediapipe/solutions/vision/face_landmarker | Apache-2.0 |
hand_landmarker.task |
7,819,105 | hand landmarks | ai.google.dev/edge/mediapipe/solutions/vision/hand_landmarker | Apache-2.0 |
mask_detector.h5 |
11,490,448 | mask | github.com/chandrikadeb7/Face-Mask-Detection | MIT |
colorization_deploy_v2.prototxt |
9,945 | colorization | github.com/richzhang/colorization | BSD-2-Clause |
colorization_release_v2.caffemodel |
128,946,764 | colorization | github.com/richzhang/colorization | BSD-2-Clause |
pts_in_hull.npy |
5,088 | colorization | github.com/richzhang/colorization | BSD-2-Clause |
haarcascade_eye.xml |
341,406 | eye color, roll alignment | github.com/opencv/opencv | Intel License Agreement (BSD-style, in the file header) |
BFM/similarity_Lm3D_all.mat |
994 | 3d reconstruction | github.com/sicxu/Deep3DFaceRecon_pytorch | MIT |
Attribution
opencv_face_detector.pbtxt: OpenCV face detector config, opencv/opencv samples/dnn/face_detector (Apache-2.0)retinaface_mobilenet0.25.onnx: RetinaFace MobileNet-0.25 by biubug6 (MIT), ONNX export by AMD (Apache-2.0)fairface_7class.onnx: FairFace (Karkkainen and Joo, WACV 2021), dchen236/FairFace, CC BY 4.0. ONNX conversion of the 7-race FairFace model; the conversion's origin is not recorded.mivolo_v2.safetensors: MiVOLO v2 (WildChlamydia/MiVOLO, iitolstykh/mivolo_v2), Apache-2.0mivolo_v2_config.json: MiVOLO v2 (WildChlamydia/MiVOLO, iitolstykh/mivolo_v2), Apache-2.0emotion_ferplus.onnx: FER+ emotion model from the ONNX Model Zoo (onnx/models), MITmini_xception_fer.h5: mini-Xception FER2013 model by Octavio Arriaga (oarriaga/face_classification), MITface_landmarker.task: MediaPipe Face Landmarker (Face Mesh V2) by Google, Apache-2.0hand_landmarker.task: MediaPipe Hand Landmarker by Google, Apache-2.0mask_detector.h5: Face mask detector by Chandrika Deb (chandrikadeb7/Face-Mask-Detection), MITcolorization_deploy_v2.prototxt: Colorful Image Colorization (Zhang, Isola, Efros, ECCV 2016), richzhang/colorization, BSD-2-Clausecolorization_release_v2.caffemodel: Colorful Image Colorization (Zhang, Isola, Efros, ECCV 2016), richzhang/colorization, BSD-2-Clausepts_in_hull.npy: Colorful Image Colorization (Zhang, Isola, Efros, ECCV 2016), richzhang/colorization, BSD-2-Clausehaarcascade_eye.xml: OpenCV Haar eye cascade, Intel License Agreement (BSD-style; full text in the file header)BFM/similarity_Lm3D_all.mat: Landmark alignment template from sicxu/Deep3DFaceRecon_pytorch (MIT)
Licenses and notices
- MIT:
retinaface_mobilenet0.25.onnx(network),emotion_ferplus.onnx,mini_xception_fer.h5,mask_detector.h5,BFM/similarity_Lm3D_all.mat - Apache-2.0:
opencv_face_detector.pbtxt,retinaface_mobilenet0.25.onnx(ONNX export),mivolo_v2.safetensors,mivolo_v2_config.json,face_landmarker.task,hand_landmarker.task - BSD-2-Clause:
colorization_deploy_v2.prototxt,colorization_release_v2.caffemodel,pts_in_hull.npy - CC BY 4.0:
fairface_7class.onnx(FairFace, Karkkainen and Joo; converted to ONNX) - Intel License Agreement (BSD-style):
haarcascade_eye.xml(full text in the file header)
LICENSE reproduces the copyright notices and full license texts these terms require.