OrientedDet pretrained weights
Checkpoints for oriented-det (pip install oriented-det). Download by slug (not by filename):
odet pretrained download oriented_rcnn_dota_le90_1x
# or in a recipe:
# "load_from_checkpoint": "hf://oriented_rcnn_dota_le90_1x"
Repo: dl4eo/oriented-det-pretrained. Docs: pretrained/README.md.
Finetune from 1×, not 3×. DOTA 3× is on Hub for AP75; leaky eval-val is train+val tile memorization. SSDD / HRSID / FAIR1M have no Hub zoo — finetune these DOTA 1× slugs locally.
DOTA eval_map50 in the package manifest is leaky val tiles. Advertised DOTA mAP50 below is official Task 1. HRSC mAP50 is held-out ImageSets test.
DOTA v1.0 le90 (official Task 1)
| Slug | Schedule | Task 1 mAP50 |
|---|---|---|
oriented_rcnn_dota_le90_1x |
1× (advertised / finetune) | 76.73% |
oriented_rcnn_dota_le90_3x |
3× (AP75) | 74.88% |
rotated_faster_rcnn_dota_le90_1x |
1× (advertised / finetune) | 74.42% |
rotated_faster_rcnn_dota_le90_3x |
3× (AP75) | 74.48% |
rotated_fcos_dota_le90_1x |
1× (advertised / finetune) | 73.07% |
rotated_fcos_dota_le90_3x |
3× (AP75) | 72.91% |
rotated_retinanet_dota_le90_1x |
1× circum-HBB | 67.87% |
rotated_retinanet_dota_le90_3x |
3× circum-HBB | 70.70% |
HRSC2016 le90 (held-out test)
| Slug | Schedule | mAP50 |
|---|---|---|
oriented_rcnn_hrsc2016_le90_3x |
3× | 90.41% |
rotated_faster_rcnn_hrsc2016_le90_3x |
3× | 88.77% |
rotated_fcos_hrsc2016_le90_3x |
3× | 88.34% |
Each slug has a content-addressed .pth plus sidecar .json / .log. Weights are Apache-2.0 OrientedDet exports. Dataset licenses (DOTA, HRSC, …) are separate; see Apache 2.0 vs DOTA / HRSC.