Open Shichang-Fu opened 10 months ago
these codes return masks while inference and you can calculate gflops manually. https://github.com/Cuogeihong/CEASC/blob/2abfd1a99f1b0fe1ed3d51588b64549e1584da50/mmdet/models/dense_heads/gfl_dy_head.py#L269-L271
Hello, author. I am trying to replicate your work. I am using mmdetection/tools/analysis_tools/get_flops.py to obtain the model's FLOPs. However, I have noticed that this function does not capture the FLOPs for DyConv2D. Could you please explain how you calculate FLOPs for sparse convolutions?Thanks![image](https://github.com/Cuogeihong/CEASC/assets/103252169/dff6252e-c473-46d4-8636-6a69806460c8)