Closed mywebinfo65536 closed 1 year ago
Yes, the structure is the same (thus the pretrained parameter can be loaded), but the operation is different. Please check https://github.com/Sense-X/UniFormer/blob/f92e423f7360b0026b83362311a4d85e448264d7/object_detection/mmdet/models/backbones/uniformer.py#L192-L210
The window block is used to reduce the computation for large-resolution input.
Okey, thanks for your answer.
I found them the same, why ?