Rotated-RetinaNet
Rotated-RetinaNet copied to clipboard
low map
dear ming:why my map is lower than your result?i follow your steps to operate on HRSC2016 dataset.what's the problem with my result? can you direct ?thanks.

hello, have you find out the way to solve the problem? i meet it too.
Sorry, I don't know what's wrong with your code without more details. I've achieved the mAP of 82+ with the following settings:
- lr=0.0002(adam),
- bs=4,
- size=800,
- epoch=100,
- dataaugment ,
- single-scale training and testing
如果还是不行可以试试这个模型:https://github.com/ming71/Cascade-Rotated-RetinaNet, 加了box refinement,对参数更加鲁棒,只要设置不太离谱可以上86+