rotcert-models / RIGHTS.md
PeterPonyu's picture
Release evaluated research models, portable configurations and parameter verification
a6303b7 verified
|
Raw History Blame Contribute Delete
3.46 kB
# Rights and attribution
Copyright 2026 Zeyu Fu for the author-trained parameter exports and author-written
release documentation, to the extent of the author's rights. The author makes
these files publicly downloadable for noncommercial research, inspection and
reproduction of the reported experiments. Preserve the scientific attribution,
parameter-verification records and the applicable third-party notices when
sharing an unchanged copy. This statement does not grant rights the author does
not hold.
The software MIT license in RotCert is **not** a blanket license for these model
weights, third-party datasets, pretrained backbones, images or upstream detector
implementations. No commercial-use permission for those third-party materials is
asserted by this release. Upstream and provider terms must be checked separately;
seek permission from the relevant rights holder where the intended use is not
covered. The model-card `other` license label points to this scope statement; it
does not assert a standardized permissive license for the whole collection.
## Detector implementations
- [MMRotate](https://github.com/open-mmlab/mmrotate), source revision
`3ff004eb21ea040455b5585db229edba4037f1bf`: the upstream code carries Apache-2.0.
Its detector implementations and any pretrained assets retain their own terms.
The portable configurations describe the evaluated author-trained models;
framework source and pretrained initialization files are not bundled here.
- [EAV-DETR](https://github.com/zzzhak/EAV-DETR), source revision
`fb99da7b0f49edffcd1fa9e45a4424e9ae86e1eb`: the inspected upstream repository did
not declare a repository license at that revision. Its code and geometry
reference are not redistributed or described as MIT-licensed here. Obtain
them from the original author and check permission for your intended use.
## Dataset providers
- [DIOR / DIOR-R](https://gcheng-nwpu.github.io/): obtain data and annotations from
the original research provider and retain its use and citation conditions.
- [DOTA-v1.0](https://captain-whu.github.io/DOTA/): obtain the images and annotations
from the dataset provider and follow its terms.
- [HRSID](https://github.com/chaozhong2010/HRSID): use the provider's dataset and
attribution requirements.
- [DroneVehicle](https://github.com/VisDrone/DroneVehicle): RGB and infrared data
remain governed by the dataset provider's terms.
- [CODrone](https://github.com/AHideoKuzeA/CODrone-A-Comprehensive-Oriented-Object-Detection-benchmark-for-UAV):
the provider states CC BY-NC-SA 4.0 and academic-research-only use. Those
attribution, noncommercial and share-alike requirements are not removed by
making this model available. This release is not permission to redistribute
the underlying images or annotations under a different license.
No dataset images, full annotation sets, proprietary fonts or third-party source
packages are distributed in this repository. Training a model and removing host
metadata do not by themselves resolve every downstream legal question. This is
an attribution and scope statement, not a warranty of unrestricted rights.
## Citation
Cite the article, the detector and dataset sources relevant to the models used,
the software version DOI where applicable, and the exact Hugging Face repository
revision plus model-file hashes. Do not describe a subsequent model release as
already included in an earlier immutable software DOI.