1. Data Augmentation Notebook has code to simulate gravitational lensing images using randomly generated lentratronamy parameters and data preprocessing  . Test/Dev/ train split.   
2. Part 1 has code to predict gravitational lensing parameters using ConvNet 
3. Part 2 has U-Net code for generating source galaxy images using a custom loss function.
As this one is large file,If you have issues with notebook open use:
https://colab.research.google.com/drive/1fbtyhdeOTflkUsTr8pbK24BaOqcmHnb6?usp=sharing

4. Experiments notebook has few good models we tried out 
As this one is also large file,If you have issues with notebook open use:
https://colab.research.google.com/drive/1NUu37FByg1v5kugLp6anZ0DIzjgOyWa5?usp=sharing 

All notebooks also available on 
https://github.com/madhumitadange/Generate-Source-Galaxy-Images-from-Strong-Gravitational-Lens-Images