Understanding people assembling through hand action recognition

Abstract

Assembly and disassembly tasks are essential to people’s daily life, its scope can range from furniture setup, appliance repair, food preparation and so on. While general hand action recognition has significantly progressed with the advances in deep learning, the specific task of understanding assembly processes remains relatively unexplored. This Final Year Project aims to develop a robust model for assembly understanding. With large datasets like Assembly101, multiple architectures are implemented for this project to further evaluate their performance with different parameters used. The performance will be evaluated using established metrics, with the goal of contributing to advancements in smart assistants and augmented reality (AR) technologies for enhanced assembly understanding.