a small mistake of resnet ,leading to inaccuracy shape of internal layer

keras-team / keras-applications

Reference implementations of popular deep learning models.

Other

2k stars 910 forks source link

a small mistake of resnet ,leading to inaccuracy shape of internal layer #161

Closed ChienWong closed 4 years ago

ChienWong commented 4 years ago

https://github.com/keras-team/keras-applications/blob/976050c468ff949bcbd9b9cf64fe1d5c81db3f3a/keras_applications/resnet50.py#L234 the MaxPooling need to add a extra argument "padding="same"", if not ,the output_shape of the layer is less 1 than normal

jaketae commented 4 years ago

I'd like to contribute to this. But before I jump in, can you clarify a bit more as to what the issue is? What do you mean by "normal" when you say "1 less than normal"?