This is the Partial Implementation of Residual Dense Network Mentioned in the Paper Residual Dense Network for Image Super-Resolution
- Mnist data set is used as the training and testing dataset
- Model defined is simple as compared to as in the paper
The 28 x 28 x 1 images were downscaled to 14 x 14 x1 for generating training and testing dataset
- The image to the left is 14 x 14 x 1 image
- The image to the right is 28 x 28 x1 image which is upscaled using Residual Dense Network
-
Note : To display gif on git , Images has been rescaled to both 400 x 400 without any alternative resizing optimizations
-
The Residual Dense Network is used to generate 28 x 28 x1 images from 14 x 14 x 1 images and, these are then upscaled to 400 x 400 x 1 without using any alternative resizing algorithms
-
One can use the hrimages.gif and lrimages.gif to compare the gifs produced by the code,to observe differences
-
low resolution images
-
High resolution images
