In the Dipanjan Session, he tweaks the vgg19 model to support 32x32 images for training. But vgg19 accepts the image size 224x224. How does TensorFlow accept such sizes? If it does accept, what kind of sizes it can take in for training?
@priya010695 while this was already responded on slack. Closing it here again for sake of completion and others whoever land here can find the solution.
The default input size for VGG19 model is 224x224. Coming to the limitation on the size, it doesn’t accept anything below 32 pixels. So technically it should support for sizes anything above 32.