SRDenseNet_paddle

This repository is implementation of the "Image Super-Resolution Using Dense Skip Connections".

Requirements

• paddlepaddle 2.4.0
• Numpy 1.15.4
• Pillow 5.4.1
• h5py 2.8.0
• tqdm 4.30.0

Train

The coco2017 50K, Set5 dataset converted to HDF5 can be downloaded from the links below.

Dataset	Scale	Type	Link
coco2017 50K	4	Train	Download
Set5	4	Eval	Download

Otherwise, you can use prepare.py to create custom dataset.

python train.py --train-file "BLAH_BLAH/coco2017_x4.h5" \
                --eval-file "BLAH_BLAH/Set5_x4.h5" \
                --outputs-dir "BLAH_BLAH/outputs" \
                --scale 4 \
                --lr 1e-4 \
                --batch-size 8 \
                --num-epochs 60 \
                --num-workers 0 \
                --seed 123                

Test

Pre-trained weights can be found in BLAH_BLAH/outputs.

python test.py --weights-file "BLAH_BLAH/outputs/x4/best.pdiparams" \
               --image-file "data/ppt3.bmp" \
               --scale 4

Results

PSNR was calculated on the Y channel.

Set5

Eval. Mat	Scale	SRDenseNet_All (Paper)
PSNR	4	32.02