/ImageSuperResolution

Image super resolution using Neural networks.

Primary LanguagePython

Image SuperResolution

Bachelor thesis on Image SuperResolution.

 

Requirements

Python version 3.9.1 was used for the development, but it is likely that any Python version above 3.7 will work as well.

Two packages from nvidia: cudatoolkit 11.2 and cuDNN 8.2.1. The versions of these two packages depend on each other and on other Python libraries.

Python libraries:

  • keras 2.6.0
  • numpy 1.19.5
  • opencv-python-headless 4.5.4.58
  • tensorflow 2.6.0
  • pillow 9.1.0

If WandB logging is to be used, the wandb package must be installed.

If Azure ML is to be used, the azureml packages must be installed.

 

Running

  • Run predict.py
  • Select image to upscale
  • Choose upscaling factor

 

Samuel Buban