Align shared attribution with published ONNX artifacts

README.md

@@ -125,5 +125,6 @@ NMS, and coordinate scaling described in the KernelVision source repository.
 - [Ultralytics YOLOv8 base model](https://huggingface.co/Ultralytics/YOLOv8)
 
 The artifacts are distributed under AGPL-3.0 in accordance with the upstream
-Ultralytics licensing terms. See the KernelVision third-party notices for
-complete attribution.
+Ultralytics licensing terms. See the
+[KernelVision third-party notices](THIRD_PARTY_NOTICES.md) for complete
+attribution.

THIRD_PARTY_NOTICES.md

@@ -13,9 +13,12 @@ License v3.0 (AGPL-3.0), subject to its published licensing terms:
 - <https://github.com/ultralytics/ultralytics>
 - <https://www.ultralytics.com/license>
 
-Ultralytics software, pretrained weights, and generated deployment artifacts
-are not included in this repository. Reproduction scripts obtain or consume
-them separately.
+The Ultralytics Python package, original pretrained PyTorch checkpoint, and
+serialized TensorRT engines are not redistributed by KernelVision.
+Reproduction scripts obtain or consume them separately. The companion
+[Hugging Face model repository](https://huggingface.co/mmoz-root/kernelvision)
+publishes two derived ONNX graphs under AGPL-3.0 with explicit provenance and
+limitations.
 
 ## Ultralytics sample image