Request for permission to use ONNX model files (best.onnx) for commercial product

[kyhv]

Hi @im-syn

Thank you for building and open-sourcing SafeVision — it's a great project.

I'm Tony, Project Manager at Rikkeisoft Corporation, Vietnam. We're interested in integrating SafeVision's content-safety detection into one of our commercial products/services.

We understand the source code is licensed under Apache-2.0, which permits commercial use. However, we noticed the ONNX model files (best.onnx, best_gender.onnx, safety_objects.onnx) are distributed separately via Releases and are not explicitly covered by a stated license.

Could you please clarify:

1. Under what license or terms the pre-trained ONNX models may be used (including commercial/production use)?
2. Whether these models were trained/exported using Ultralytics' YOLO framework or codebase — and if so, whether Ultralytics' AGPL-3.0 license (or their commercial Enterprise license) applies to the model weights themselves, separate from the Apache-2.0 license on the repo's source code.
3. Whether commercial use of the models requires separate permission, attribution, or a paid license.
4. If there are any restrictions on modifying, fine-tuning, or redistributing the models as part of our product.

We'd be happy to credit the project and comply with any conditions you require. Please let us know if a written confirmation (e.g., via email) would work better for you — happy to continue there.

Thank you very much for your time and for maintaining this project.

Best regards,
Tony — Rikkeisoft Corporation