NOTICE: WEIGHTS LICENSING

No pretrained weight files are distributed in this directory or anywhere
else in the source tree. The conversion code (`coreyolo convert` /
`coreyolo/export/convert.py`) is part of the CoreYOLO source repository
and is covered by the root MIT License.

Published checkpoints belong on GitHub Releases. The catalog is
`weights/manifest.json`. Each row states `weights_license` for that
file. Converting a checkpoint does not change its applicable terms.
An ONNX or Core ML artifact built from a restricted checkpoint inherits
the restriction. Choosing a model means choosing its license.

Per-source summary:

 GELAN COCO (coreyolo-{n,s,m,l}-coco) MIT
 upstream: MultimediaTechLab/YOLO v1.0-alpha
 files: v9-t.pt → n, v9-s.pt → s, v9-m.pt → m, v9-c.pt → l
 Copyright (c) 2024 Kin-Yiu Wong and Hao-Tang Tsui
 The authors' MIT re-release, not WongKinYiu/yolov9 (GPL-3.0) and
 not Ultralytics yolov9{t,s,m,c}.pt (AGPL-3.0).
 Convert with: coreyolo convert --weights v9-t.pt
 Keep the MIT copyright and permission notice with any copy.

 GELAN / E2E / segment trained from scratch MIT
 upstream: CoreYOLO train (`coreyolo train` or Detector.train)
 Native recipes (coco-n, coco-s, coco-m, coco-l, coco-n-e2e,
 segment) are original weights. ReLU, Neural Engine.

 Ultralytics yolov9{t,s,m,c}.pt AGPL-3.0
 Those files inherit AGPL-3.0 from their source. Convert still
 remaps names, but the checkpoint is stamped AGPL-3.0 and is NOT
 covered by the MIT License that applies to CoreYOLO code.
 Do not rehost them as MIT.

The Hugging Face / GitHub Release page of the exact file you download
is authoritative, and it can change without this notice changing with
it. See docs/licenses.md. This file is a description, not legal advice.
