This repository is unofficial implementation of Stacked Conditional Generative Adversarial Networks for Jointly Learning Shadow Detection and Shadow Removal [Wang+, CVPR 2018] with PyTorch.
Official Dataset and Code(coming soon...) is here.
./dataset
. You can Download datasets from here. Then,
python3 train.py
When Testing images from ISTD dataset.
python3 test.py -l <checkpoint number>
When you would like to test your own image.
python3 test.py -l <checkpoint number> -i <image_path> -o <out_path>
Here is a result from test sets. (Left to right: input, ground truth, shadow removal, ground truth shadow, shadow detection)
Here are some results from validation set. (Top to bottom: ground truth, shadow detection)
Here are some results from validation set. (Top to bottom: input, ground truth, shadow removal)
You can download from here.