Thank you very much to the author for sharing excellent ideas and code, but I have a little problem. In the SRA description part of the paper, in formula 3, the channel is changed by matrix multiplication after reshape first, but in the code, it is realized by a convolution. The complexity of the two is different. The author has experimented in two ways Is the result similar?Looking forward to your answer.
Thank you very much to the author for sharing excellent ideas and code, but I have a little problem. In the SRA description part of the paper, in formula 3, the channel is changed by matrix multiplication after reshape first, but in the code, it is realized by a convolution. The complexity of the two is different. The author has experimented in two ways Is the result similar?Looking forward to your answer.