Non-local_pytorch icon indicating copy to clipboard operation
Non-local_pytorch copied to clipboard

implement in other network

Open sxzy opened this issue 6 years ago • 2 comments

SO if I want to add nonblock into my own experiment I can only add NONLocalBlock2D() into somewhere of the network ? and I did't need to change other params ?

sxzy avatar Oct 29 '18 09:10 sxzy

Yes, but the size of feature map should not be too small, such as 1x1, 2x2.

AlexHex7 avatar Oct 31 '18 00:10 AlexHex7

Yes, but the size of feature map should not be too small, such as 1x1, 2x2.

感谢提供的代码!想请教一个问题,我在孪生神经网络两个分支分别加了Non-local模块,其中一个分支的feature map 大小是55512,只要在这个后面加Non-local模块,效果就非常差,另外一个分支是2121512,就没有问题,是因为尺寸太小了吗?请问为什么尺寸小效果会不好呢?

lc0716 avatar May 06 '21 06:05 lc0716