lucciffer/flare_removal_pytorch

PyTorch Implementation of "How to Train Neural Networks for Flare Removal"

Usage

installation

after install torch and torchvision

pip install loguru fire omegaconf tqdm scikit-image kornia opencv-python

training

python3 main.py ./configs/original_paper.yml --gpus 0
# change loss weights in original_paper.yml
python3 main.py ./configs/original_paper.yml --gpus 0 name="ffl_1-lpips_1|l1_1-lpips_1" \
    loss.weight.flare.ffl=100 \
    loss.weight.flare.l1=0 \
    loss.weight.flare.lpips=1 \
    loss.weight.flare.perceptual=0 \
    loss.weight.scene.ffl=0 \
    loss.weight.scene.l1=1 \
    
    loss.weight.scene.lpips=1 \
    loss.weight.scene.perceptual=0 \

evaluate using paired images

python3 test.py evaluate </path/to/config.yml> </path/to/checkpoint>

generate non-flare images

python3 test.py generate </path/to/config.yml> </path/to/checkpoint> \
    --image_folder </path/to/images-with-flare/> \
    --output_folder </path/to/pred-images/>