Skip to content

THIRD_PARTY_NOTICES.md

Latest commit

 

History

History
154 lines (108 loc) · 6.07 KB

THIRD_PARTY_NOTICES.md

File metadata and controls

154 lines (108 loc) · 6.07 KB

Third-Party Notices (OBF Vision)

This repository ("OBF Vision") is licensed under the GNU Affero General Public License v3.0 (AGPL-3.0). This file lists third-party software and externally downloaded assets that OBF Vision can use.

Important:

  • Third-party components remain under their respective licenses.
  • This repository is intended to contain source code only. Model weights, biometric data, and runtime artifacts SHOULD NOT be committed.
  • If you redistribute a packaged/binary distribution that bundles third-party components, you may need to include additional license texts and notices (see “Bundled binaries” below).

A) Direct Python dependencies (from requirements.txt)

The project depends on the following Python packages (installed via pip). They are NOT vendored into this repository.

  1. faiss-cpu

    • Purpose: Vector similarity search (FaceDB index).
    • Upstream: Faiss (Meta).
    • License: MIT (Faiss). Packaging metadata may reference additional permissive licenses.

  2. fastapi

    • Purpose: Web API framework.
    • License: MIT.

  3. numpy

    • Purpose: Numerical computing / arrays.
    • License: BSD-3-Clause (NumPy) + additional permissive licenses for bundled components.

  4. opencv-python

    • Purpose: Image processing (cv2).
    • License (package build scripts): MIT.
    • OpenCV library itself: Apache-2.0 (for OpenCV >= 4.5.0).

  5. openvino

    • Purpose: Inference runtime / model conversion APIs.
    • License: Apache-2.0.

  6. pillow

    • Purpose: Image I/O fallback.
    • License: MIT-CMU (Pillow / PIL fork).

  7. python-multipart

    • Purpose: Multipart/form-data parsing (uploads).
    • License: Apache-2.0.

  8. ultralytics

    • Purpose: YOLO export/conversion to OpenVINO (used during model install jobs).
    • License: AGPL-3.0 (Ultralytics provides enterprise/commercial licensing options separately).

  9. uvicorn

    • Purpose: ASGI server.
    • License: BSD-3-Clause.

  10. websockets

  • Purpose: WebSocket support.
  • License: BSD-3-Clause.

B) Additional runtime dependencies (imported, not pinned in requirements.txt)

These are imported by the codebase and typically arrive as transitive dependencies or must be installed explicitly.

  • pydantic

    • Purpose: Configuration schema/validation.
    • License: MIT.

  • torch (PyTorch) (OPTIONAL / required for some conversion flows)

    • Purpose: Used indirectly by Ultralytics export, and referenced in ONNX conversion workaround.
    • License: BSD-3-Clause (PyTorch).

  • Open Model Zoo tools (OPTIONAL; required for omz_downloader / omz_converter)

    • Purpose: Download/convert certain Intel Open Model Zoo models (e.g., face-reidentification-retail-0095).
    • Typical install: via openvino-dev / OMZ tools.
    • License: Apache-2.0 (Open Model Zoo).

C) Externally downloaded model files (NOT included in this repo)

OBF Vision can download model weights from third-party sources at runtime (Model Manager). These files are not part of this repository and must not be committed.

C1) Ultralytics pretrained weights (.pt) downloaded from GitHub Releases

Examples configured in MODEL_CATALOG:

  • yolo11n / yolo11s / yolo11m (.pt) from Ultralytics assets releases
  • yolov8n / yolov8s (.pt) from Ultralytics assets releases

Notes:

  • These are provided by Ultralytics as part of their model distribution ecosystem.
  • License and usage terms depend on Ultralytics’ licensing (AGPL-3.0 vs enterprise/commercial).
  • Do not redistribute these weights unless you are sure you have the rights to do so.

C2) yolov8n-face face detector (Google Drive link)

Configured source:

  • Google Drive direct download URL (ID: 1qcr9DbgsX3ryrz2uU8w4Xm3cOrRywXqb)

Notes:

  • The license/rights of this specific file are not determined by this repository.
  • Treat as “downloaded third-party asset” and do not redistribute unless you have explicit permission.

C3) InsightFace-derived ONNX ReID model (w600k_mbf.onnx) from Hugging Face

Configured source:

  • Hugging Face: deepghs/insightface (buffalo_s/w600k_mbf.onnx)
  • Hugging Face lists a custom “model-distribution-disclaimer-license”.

Important:

  • InsightFace upstream notes that while the code is MIT, training data and models trained with those data may be restricted (e.g., non-commercial research only) depending on the specific model/data provenance.
  • You are responsible for reviewing and complying with the model’s license and any upstream restrictions before using it, especially for commercial use.

Recommendation:

  • Keep this model out of the repository.
  • Consider switching to a model with a clear permissive license for your intended use.

C4) Intel Open Model Zoo models

Example configured source:

  • face-reidentification-retail-0095 via omz_downloader / omz_converter

Notes:

  • Open Model Zoo is licensed under Apache-2.0.
  • Individual model directories can include additional notices; check OMZ model documentation if needed.
  • Do not commit downloaded models; download on demand.

D) Bundled binaries and platform-specific notes

This repository does not ship binary wheels, but users install dependencies via pip.

If you redistribute a packaged application that bundles dependencies (e.g., frozen executable / Docker image with wheels), you may need to provide additional license texts/notices for bundled components.

Examples:

  • opencv-python wheels may bundle third-party components (e.g., FFmpeg and, on some platforms, Qt) with their own licenses.
  • Always review the specific wheels you distribute and their “3rd party licenses” files (if present).

E) Trademarks

Intel, OpenVINO, and other marks are trademarks of their respective owners. This project is not affiliated with or endorsed by the upstream vendors unless explicitly stated.

F) Where to find license texts

  • Python package licenses: see the package’s PyPI page and upstream repository LICENSE file.
  • Ultralytics licensing: see Ultralytics official license information and repository license.
  • OpenVINO / Open Model Zoo: see Intel’s repositories and license files.
  • InsightFace: see upstream license policy and model distribution terms.

THIS FILE IS GENERATED BY AI