How does OctConv really implementstrided convolutions
rongmelt opened this issue · 1 comments
rongmelt commented
First of all, feel the author's open source spirit, great idea!
I know that low and high frequencies are divided according to a given proportion, but how are they divided along the channel dimension? Are they divided randomly according to proportion? Or are there other pre-processing in it?
Thank you very much.
marnie007 commented
从代码上看,感觉高频的是通过一层卷积后得到,低频的话,是先pooling 再卷积得到的