The results on the HRSC2016 dataset are very strange

[zhuyh1223]

Thank you for sharing your work. When I evaluated TIOE on HRSC2016, I got 0.49 and 0.38 AP50 using Yolov5m and Yolov5x. The results are so strange. I strictly followed the steps you wrote in readme.md to conduct the experiment. The batch size was set to 8 and one GPU was used. Can you help me analyze the reason or provide the checkpoints of your training on HRSC2016?

[ming71]

Did you implement the method yourself, or use the framework in this repo?