Confusing convnet demos in paddlepaddle book's ch02 and ch03
Created by: Meleeri
Hi, I'm facing some difficulty in figuring out how paddle handles cnn convnets. According to demos in paddlepaddle book's ch02 and ch03, all inputs are flattened before fed to the network as the data layer, which is very confusing since no width and height are specified, convolution seems impossible. Is there something missing in the documentation or paddle simply treats inputs as squares when no width and height are specified?