- Download restructured RealSR version 2 for training:
python download.py --file RealSR_234
- (Optional) To download generated dataset with intermidiate levels from Interflow:
python download.py --file RealSR_24_gen
- To train SR networks, run
python3 main.py \
--GT-paths [TRAIN DATASET PATH] \
--parameter-name [FOLDER NAME TO SAVE THE RESULTS] \
--train-iteration [NUM TRAIN ITERATION] \
--validation-step [VAL ITERATION] \
--test-scale 3.0 \
--model [VDSR|RCAN|KPN|HAN|NLSN|SwinIR|MetaSR|LIIF] \
--gpu [GPU INDEX];
Please refer more examples in scripts/train.sh
- Download the pre-trained model and place it in
./pretrained_models/
manually
OR run:
python download.py --file pretrained
- To download RealSR x3 test datasets[link], run
python download.py --file RealSR_234
- To download DRealSR x3 test datasets [offical link], run
python download.py --file DRealSR_3_test
- Testing
python3 main.py \
--parameter-name [FOLDER NAME TO SAVE THE RESULTS] \
--parameter-restore-path [PATH TO YOUR CHECKPOINT] \
--test-scale 3.0 \
--model [VDSR|RCAN|KPN|HAN|NLSN|SwinIR|MetaSR|LIIF] \
--test-data [RealSR|DRealSR] \
--gpu [GPU INDEX] \
--test;
Please refer more examples in scripts/test.sh