Ming Qi
Ming Qi
UCAS-ADO数据的处理放在这个repo了: [UCAS-AOD-benchmark](https://github.com/ming71/UCAS-AOD-benchmark).直接下载本地跑一下就行
我没用过大于1的，应该没问题你可以试试
Hi, I've provided the related codes, you can execute `python datasets/UCAS_AOD/ucas_aod2gt.py` to convert to dst format.
check dataset convert tools [here](https://github.com/ming71/toolbox). I'll support convert more datasets to DOTA format when I am free.
> there is no trained models in your connection,where can I find it? I've uploaded the trained models before, please dive into the document.
这是yolov3的角点坐标偏移，建议结合论文阅读
https://pjreddie.com/media/files/papers/YOLOv3.pdf
修改一下utils/dataset.py文件的标签读取函数就行
First, generate anchor settings via `utils/kmeans.py`, and you'll obtain infomationfor anchor shapes in `anchor-cluster.txt`. Next copy them and paste into `cfg/yolov3.cfg` like [this](https://github.com/ming71/yolov3-polygon/blob/4a141c503bc46ca6148bb3f60f293b23edd2d996/cfg/yolov3.cfg#L609).