Jiaming Han
Jiaming Han
We remove the 'optimizer' params using [tools/publish_model.py](https://github.com/csuhan/s2anet/blob/master/tools/publish_model.py)
https://github.com/csuhan/s2anet/blob/master/configs/rotated_iou/README.md
https://www.kaggle.com/guofeng/hrsc2016
添加 'container-crane'： https://github.com/csuhan/s2anet/blob/d2870963815b10c8f7796577e30b7c3b14e47ac6/DOTA_devkit/convert_dota_to_mmdet.py#L10-L14
Maybe you can add 'EQLv2' to this line so that we do not use another sampler. https://github.com/csuhan/s2anet/blob/d2870963815b10c8f7796577e30b7c3b14e47ac6/mmdet/models/anchor_heads_rotated/s2anet_head.py#L64
See [DOTA_devkit/convert_dota_to_mmdet.py](https://github.com/csuhan/s2anet/blob/master/DOTA_devkit/convert_dota_to_mmdet.py)
理论上，先检测再检索的模型效果更好
参考：https://github.com/csuhan/s2anet/blob/7d66620f02b29e9008fbdb91cfcb10916ea47c49/tools/test.py#L37-L57
https://github.com/open-mmlab/mmdetection/issues/1223
我们主要采用ORPooling 