BangguWu / ECANet

Code for ECA-Net: Efficient Channel Attention for Deep Convolutional Neural Networks
MIT License
1.24k stars 197 forks source link

About paramters and Flops? #35

Closed dongsheng1991 closed 3 years ago

dongsheng1991 commented 3 years ago

Why are the parameters and calculations of the released code different from those reported in your paper? (resnet50 parameters: 25.56M vs 24.37M; Flops: 4.12G vs 3.86G)

BangguWu commented 3 years ago

Thanks for your question. Different papers may use different conversion units when calculating the amount of parameters and calculations. The ECA-Net uses 1024.