This repository holds all the codes written by me during my PhD. Each sub-folder represents a chapter as follows:
This chapter lays the foundation for Neural Ordinary Differential Equations (ODEs). It introduces my innovative neural ODE model, specifically designed for dense prediction tasks, detailing its development and potential applications.
Here, I delve into the intricacies of James-Stein estimators and normalization layers. I present my pioneering method, JSNorm, which represents a significant advancement in this field.
This chapter discusses a groundbreaking method in image restoration, with a particular focus on its application in answer grounding. The novel approach outlined here marks a significant contribution to the domain.
In this chapter, I introduce the Embedding Attention Blocks (EABs), a key innovation that has achieved state-of-the-art accuracy in answer grounding. The chapter details the development and effectiveness of these blocks.
The focus here is on a comparative analysis of Apple Live Photos and Android Motion Photos versus static images, particularly in their utility for visual assisting applications. This chapter highlights the superior performance of Live/Motion Photos in common tasks.