
MIT License

Copyright (c) 2022, Jin Hyun Cheong, Tiankang Xie, Sophie Byrne, Eshin Jolly, Luke Chang

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

Scope and third-party notices
=============================

The MIT grant above covers Py-Feat's original software and documentation.
Third-party source code retains its own notices. Downloaded pretrained
weights, fitted statistical parameters, datasets, and example media have
separate provenance and terms; this file does not grant rights held by
their providers or revoke existing valid grants.

Separate model-family notices and dataset evidence:

- Classic Py-Feat v1.0 / Detectorv1: LICENSES/PYFEAT-V1.md
- Multitask Py-Feat v2.0 / Detectorv2: LICENSES/PYFEAT-V2.md
- PyFeat-Live application: LICENSES/PYFEAT-LIVE.md
- Training and evaluation datasets: LICENSES/DATASETS.md

An MIT implementation license does not establish every checkpoint's
commercial-use or redistribution permissions. Nor does a dataset's
research-only designation automatically license all resulting weights.
The model-family notices identify explicit restrictions and unresolved
permission scope, including cases where a research-only label is not
sufficient to authorize public distribution.

Upstream references (their scope must be read in the linked materials):

- RetinaFace: https://github.com/biubug6/Pytorch_Retinaface/blob/master/LICENSE.MIT
- FaceBoxes: https://github.com/cleardusk/3DDFA_V2/blob/master/LICENSE
- MTCNN: https://github.com/ipazc/mtcnn/blob/master/LICENSE
- Img2Pose: https://github.com/vitoralbiero/img2pose/blob/main/license.md
- Mobilenet: https://github.com/cunjian/pytorch_face_landmark (basenet)
- MobileFaceNet: https://github.com/cunjian/pytorch_face_landmark (mobilefacenet)
- PFLD: https://github.com/cunjian/pytorch_face_landmark (pfld_compressed)
- Residual Masking Network: https://github.com/phamquiluan/ResidualMaskingNetwork/blob/master/LICENSE
- Facenet: https://github.com/timesler/facenet-pytorch/blob/master/LICENSE.md
- ArcFace / InsightFace (code): https://github.com/deepinsight/insightface/blob/master/LICENSE (MIT)
- ArcFace ResNet50 pretrained weights (`buffalo_l`'s `w600k_r50.onnx`,
  repackaged as `arcface_r50.safetensors` at
  https://huggingface.co/py-feat/arcface_r50): distributed under
  InsightFace's non-commercial-research terms. The underlying training
  data (WebFace600K, a Tsinghua-curated subset of WebFace260M:
  https://www.face-benchmark.org/) is also non-commercial-research only.
  Py-Feat's conversion code and integration are MIT-licensed, but the
  converted weights inherit the upstream restriction. Commercial users
  must independently validate license compatibility for their use case
  and may need to substitute a different identity model.
