CODEWITHSUNDEEP
pythonneural-networkgenerative-adversarial-network
Arihant Kamdar
Arihant Kamdar

Reputation: 124

GANs output is not optimum

I am trying to implement GANs for super-resolution enhancement using Tensorflow. Here is the link of my collab code https://colab.research.google.com/drive/1F376Y6T5ehoE-mk2q7LGBmLq-OkWxqtr?usp=sharing The ground truth iamge

the expected image

First image is the output image

Second is the Expected output

Can someone help me figure out what is wrong in the image. Any help would be appreciated. Thank you

Upvotes: 0

Views: 63

Answers (1)

LazyAnalyst
LazyAnalyst

Reputation: 476

Actually, your generator looks quite simple. Try adding some more convolutional layers Conv2D after upsampling the inputs. There are also many techniques & architectures (such as Pix2Pix) that handle very well high quality images.

This repository contains quite a few GANs that might help you improve your model + some tricks You can do to improve the overall quality: https://github.com/kochlisGit/Keras-GAN

Upvotes: 1

Related Questions