Enhanced notes capture using super resolution technique on tecogan

Lee, Theresa Wen Yan (2023) Enhanced notes capture using super resolution technique on tecogan. Final Year Project, UTAR.

Preview

PDF Download (108Mb) | Preview

Abstract

This project is an Artificial Intelligence (AI) training project for the images or notes resolution enhancement for academic purpose. The concept is similar to documents scanning using mobile devices which users are able to capture the messages on the whiteboard, screen and notebook, then images can be further saved as enhanced notes in a softcopy file. The study areas included machine learning as well as generative adversarial network (GAN). The work that will be proposed in this project will be enhancing notes captured using the super resolution technique on TecoGAN. Super resolution imaging technique is a technique which the missing finegrained in the low-resolution images will be filled up and further enhanced into a highresolution image. Adversarial training has proven to be a quite effective method in images super-resolution. By using generative adversarial network (GAN), the quality of the visual perception can be greatly improved. TecoGAN, a new Artificial Intelligence (AI) Super-Resolution Algorithm will be introduced in this project. An image super-resolution model will be trained using TecoGAN model to improve the clarity of blurry photos that captured from a whiteboard in the lecture classroom. Several tools such as Tensorflow, PyTorch, JupyterNotebook and Anaconda will be used in the project development. The outcome of this project will be the high resolution of blurry long-distance images which captured by the standard mobile phone camera of students or users, enhanced by using the TecoGAN model.

Actions (login required)