why the code of resnet is different from kaiming He paper

[verigle]

The bottleneck in resnet50,101 is consist of 1x1, 3×3,1×1，but why you use two 3×3 instead it in your code? The sequential of bottleneck is also different, you use BN ->conv->prelu->BN, but in offical implementation is BN->relu->conv.

[ruiming46zrm]

@verigle did you get the answer?

[verigle]

not get the answer