I built the bellow convolution autoencoder and trying to tune it to get encoder output shape (x_encoder) of [NxHxW] = 1024 without increasing loss. Currently my output shape
