search

tensorlayer / SRGAN

Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network
https://github.com/tensorlayer/tensorlayerx
3.32k stars 813 forks source link

Main requirement is to get image deblur and denoising. Which GAN model is suitable for this requirement. #128

Open HARIHARAN1103 opened 5 years ago

HARIHARAN1103 commented 5 years ago

I have implemented SRGAN in google colab. I got only larger image, but the qualities are not improved. To get deblur,denoising and clear image what model GAN is suitable.

mexicantexan commented 5 years ago

This network takes a high-resolution photo, downgrades it to a lower resolution and works to bring that lower resolution photo back up to the same quality and standard as the high-resolution photo. Therefore, if you train on bad images, the model will perform poorly. Also if your images already has extreme amounts of noise, this network will not perform the best either. To de-noise a picture you could use a simpler GAN that takes good images, adds random noise, and then tries to get that grainy image back to the original quality.