我在使用自己的数据集进行rtmdet-inst模型训练和推理的时候出现了问题
I have a problem with rtmdet-inst model training and inference using my own dataset!
训练日志显示box的map是正常的，但是segm的map一直为0
The training log shows that box's map is normal, but segm's map is always 0
2023/12/26 20:49:42 - mmengine - INFO - Saving checkpoint at 200 epochs
2023/12/26 20:49:47 - mmengine - INFO - Evaluating bbox...
2023/12/26 20:49:48 - mmengine - INFO - bbox_mAP_copypaste: 0.987 1.000 1.000 -1.000 0.987 -1.000
2023/12/26 20:49:48 - mmengine - INFO - Evaluating segm...
2023/12/26 20:49:48 - mmengine - INFO - segm_mAP_copypaste: 0.000 0.000 0.000 -1.000 0.000 -1.000
此外我在进行推理的时候碰到了如下报错
Additionally I encountered the following error when reasoning about it
AssertionError: binary_marks must have the same shape with image