Anomaly detection in industrial settings

• Main Author: Yeo, Nathaniel Chong Ho
• Other Authors: Lin Guosheng
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
• Online Access: https://hdl.handle.net/10356/172674
• Institution: Nanyang Technological University

Anomaly detection (AD) and anomaly segmentation (AS) is a large part of various industrial setting such as the manufacturing industry. The improvement AD/AS methods could potentially save companies cost and improve the overall workflow. In recent years, AD/AS methodology with the use of computer vision (CV) has drastically improved with some models reaching near perfect accuracy in the MVTec dataset. However, most of these models have been trained on a large dataset which may not be feasible for some industries. In this paper, we aim to breach the gap and propose a training paradigm that will allow for preexisting AD/AS methodology to be applied in such scenarios.