Thermal Images Super-resolution Challenge
requirements:
- timm
- tfrecord
SwinIR + WGAN, use L1Loss & TVLoss & SSIMLoss
python make_dataset.py
# if you want to see the images in tfrecord files
# python -W ignore test_data.py
# for paired x4
sh train.sh
# for paired x2, add --scale 2 to train.sh
# for unpaired x2
sh train_cycle.sh
coming soon...
SwinIR x4: Google Drive or Baidu Pad p: wcl1
put it in "./models/x4", then python test.py
test on ubuntu20.04LTS pytorch-1.7.0 RTX3090