Single Image Super Resolution

Project Introduction

My mentor is Taegyun Jeon.

The task is to generate high resolution image from a single image. I will
study this topic from different perspectives, including both traditional and
deep method. I will apply some effective technologies on this task,
such as residual learning, various GANs, Pixel CNN, the attention mechanism in sentence processing.

Some Applications

• Photography

• De-Mosaic

Goal

• implement a complete library based on tensorflow for single image super resolution, including data preprocessing, model train, demo, evaluation.

• collect some papers with good idea and implement these models in our library, now we have a project in github named tf-sr-zoo.

• Optimize the neural network models, from two perspectives, weight storage and computation speed, making it possible to running on some mobile devices.

If you want to know more details, you can visit
my personal page (bingzhewu.github.io)