Extension of the Gated-Pixel CNN architecture (https://arxiv.org/abs/1606.05328) to removal of clouds from aerial imagery
For training on the CIFAR-10 dataset, execute
qlua run.lua -b <batch-size> -r <initial learning rate> -o <save directory path> -m <momentum, if used>
Requirements:
Torch, Lua, nn, nngraph modules for Torch
Look under data.lua to change the number of samples used
Extension to clouds is under construction (via adaptive masking)