This repo is the homebase of a community driven course on Computer Vision with Neural Networks. Feel free to join us on the Hugging Face discord: hf.co/join/discord
The implementation of the VGG19 had a bug due to missing convolutional layers, making the classifier input volume shape to be 14x14x512 instead of 7x7x512.
The implemented architecture could be called VGG15 by setting the classifier input layer to nn.Linear(14x14x512, 4096).
I have changed the code to make VGG19 w.r.t the original paper.
The implementation of the VGG19 had a bug due to missing convolutional layers, making the classifier input volume shape to be 14x14x512 instead of 7x7x512.
The implemented architecture could be called VGG15 by setting the classifier input layer to
nn.Linear(14x14x512, 4096)
.I have changed the code to make VGG19 w.r.t the original paper.