Licenses

This repository mirrors third-party models. Each file remains under its upstream license, reproduced/linked below. Arca claims no ownership of any model here.

arcface.onnx — Apache-2.0

Source: https://github.com/onnx/models/tree/main/validated/vision/body_analysis/arcface (arcfaceresnet100-8.onnx), from the ONNX Model Zoo. ArcFace is by Deng et al. Licensed Apache-2.0 per the model's directory in the ONNX Model Zoo.

auraface.mlpackage.zip — Apache-2.0

Source: https://huggingface.co/fal/AuraFace-v1 (glintr100.onnx), converted to Core ML. fal/AuraFace-v1's LICENSE.md was fetched and read in full: it is stock Apache-2.0 with no field-of-use limit, no custom attribution requirement, and no acceptable-use rider — permissive, matching the ONNX Model Zoo files above. The model card states it was "trained on commercially and publicly available data sources to enable its usage in commercial setting."

Deliberately only glintr100.onnx from that repo. fal/AuraFace-v1 also ships scrfd_10g_bnkps.onnx, 2d106det.onnx, 1k3d68.onnx, and genderage.onnx — stock InsightFace detection/landmark/attribute models whose upstream terms restrict the pretrained weights to non-commercial research (InsightFace's own repo: "the training data containing the annotation (and the models trained with these data) are available for non-commercial research purposes only"). None of those are needed or mirrored: Arca gets face detection and landmarks from the platform's own Vision/FaceAnalysis API on every platform, never from an ONNX/CoreML detector. glintr100.onnx — the recognition/embedding model — is the one file whose license this repo actually verified end to end.

clip-vit-b32-image.onnx / clip-vit-b32-text.onnx — MIT

Sources: https://huggingface.co/Qdrant/clip-ViT-B-32-vision and https://huggingface.co/Qdrant/clip-ViT-B-32-text. ONNX exports of OpenAI's CLIP ViT-B/32. Both upstream repos declare license: mit. Original CLIP is by OpenAI, released under MIT.

bpe_simple_vocab_16e6.txt.gz — MIT

Source: https://github.com/openai/CLIP (clip/bpe_simple_vocab_16e6.txt.gz). OpenAI's CLIP repository is MIT-licensed.

Deliberately NOT hosted here

Ultralytics YOLO — v5, v8, v11 and the wider family — is AGPL-3.0. AGPL requires that a larger work embedding these weights be released in full under AGPL-3.0, or that the user hold an Ultralytics Enterprise License. Arca is closed-source, so these weights are neither shipped nor mirrored. See https://www.ultralytics.com/license.

Two things worth stating plainly, because both are easy to get wrong:

Putting the feature behind a paywall does not help. AGPL triggers on distribution, regardless of whether the recipient paid.
Wrapping it in a service does not launder it. CodeProject.AI's object-detection module is itself AGPL-3.0 and built on Ultralytics; reaching the same weights over HTTP carries the same obligation. If anything AGPL is stricter about network use — that is what the extra "A" is for.

If Arca ever needs object detection, the model must be Apache-2.0 (YOLOX from Megvii, RT-DETR, or RTMDet — note YOLOX shares the "YOLO" name with Ultralytics but is an unrelated lineage and a different license).

Previously hosted, removed

yunet.onnx — MIT, from opencv/opencv_zoo (face_detection_yunet_2023mar.onnx), by Shiqi Yu and Wei Wu. Removed: Arca detects faces with the OS FaceDetector, so it was never loaded.
yolox_tiny.onnx — Apache-2.0, the official 0.1.1rc0 export from Megvii-BaseDetection/YOLOX. Removed: mirrored ahead of a feature that doesn't exist yet.