Tensorflow implementation of ESPCN algorithm described in [1]. This project was done during the Google Summer of Code 2019 program with OpenCV [2].

To run the training:

1. Download training dataset (DIV2K [3])
   bash download_trainds.sh
2. Run the training for 3X scaling factor
   python main.py --train --scale 3
   or
   Set training images directory
   python main.py --train --scale 3 --traindir /path/to/dir

To run the test:
python3 main.py --test --scale 3
python3 main.py --test --scale 3 --testimg /path/to/image

To export file to .pb format:

1. Run the export script
   python3 main.py --export --scale 3

There are trained .pb files in the export folder, for 2x, 3x and 4x scaling factors.

Example:
(1) Original picture
(2) Bicubic scaled (3x) image
(3) ESPCN scaled (3x) image

References

[1] Shi, W., Caballero, J., Huszár, F., Totz, J., Aitken, A., Bishop, R., Rueckert, D. and Wang, Z. (2019). Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network. Available at: https://arxiv.org/abs/1609.05158
[2] https://summerofcode.withgoogle.com/projects/#4689224954019840
[3] Agustsson, E., Timofte, R. (2017). NTIRE 2017 Challenge on Single Image Super-Resolution: Dataset and Study. Available at: http://www.vision.ee.ethz.ch/~timofter/publications/Agustsson-CVPRW-2017.pdf
https://data.vision.ee.ethz.ch/cvl/DIV2K/