EverythingGANs

In this project, I create several Generative Adversatial Network models for various Image to Image translation tasks. The models were first trained using tensorflow. Then, I use Flask to create a web-based implementation for uploading images and getting the augmented image.

Motivation

After watching a video about CycleGAN on Two Minute Papers I was immediately fascinated with the idea of GANs. Mingled with a desire to improve my computer vision knowledge as well as my ability to read research papers and implement them, I decided to create this project. Not only have I grown more comfortable in reading research papers(which at one time looked scary math jiggerish!), I have also had a lot of fun with this project.All in all this project served as way to :

• Help me get comfortable with reading Research Papers
• Increase my Computer Vision knowledge
• Improve my tensorflow coding skills

Demo

Architectures used

• Image Deraining - Implemented Pix2Pix architecure (Paper)
• Animate Me - Trainined CycleGAN model (Paper) on the selfie to anime dataset used in U-GAT-IT paper (Github Repo)
• Image Superresolution - Used SRGAN architecture (Paper) and used the Div2k dataset in Tensorflow Datasets

---

GANs are lovely, dark and deep,but I have no gpu to keep and a lot of epochs to run before I sleep.

-- Me frustated and tired of training GANs