Skip to content

Latest commit

 

History

History
14 lines (10 loc) · 646 Bytes

README.md

File metadata and controls

14 lines (10 loc) · 646 Bytes

4x Super-resolution by CNN

Investigation in 4x Super-resolution by Deep Convolutional Neural Networks. The network designs are my own and they work in the RGB color space. The concepts are taken from the papers in the "Papers" directory.

Data: Saved at Kaggle: https://www.kaggle.com/snirjhar/superresolution-dataset-5k

Investigations regarding 3 major factors are carried out.

  • Incremental upsampling
  • Perceptual Loss
  • Sup-pixel convolution

For results and more info see the Powerpoint Presentation.