SRCNN is a deep convolutional neural network that learns end-to-end mapping of low resolution to high resolution images.
The aim of the project is to deploy SRCNN using keras. Here is the original paper: Image Super-Resolution Using Deep Convolutional Networks. The implementation is found in the notebook srcnn.ipynb
Metrics used for evaluation:
- Peak Signal to Noise Ratio (PSNR)
- Mean Squared Error
- Structural Similarity Index
Data Used: I've used the same data as mentioned in the paper. You can download it from here. Copy both the Set5 and Set14 datasets into a new folder called 'source'.
Preparing data:
- Produce low resolution version of the images through bilinear interpolation.
- Save the new images in a folder.
The achitecture and hyper parameters necessary to build the SRCNN network can be obtained from the publication referenced above.
Some Results:
Future plans for this project: Compare all super resolution models including real-time video enhancement. Currently working on implementing the new SR3 paper. The results will be updated soon.