Publication Date

Fall 2024

Degree Type

Thesis

Degree Name

Master of Science (MS)

Department

Computer Engineering

Advisor

Kaikai Liu; Haonan Wang; Wencen Wu

Abstract

For over a decade, computer vision (CV) has become an indispensable component of numerous camera surveillance applications as well as intelligent autonomous systems. Thanks to new advances in AI, Edge Computing, and IoT technologies, there is now a rapidly growing number of smart camera devices, industrial and consumer robots that are using computer vision for various applications - from human tracking and analysis, object classification, to visual inspection and anomaly detection. For most common use cases, building vision-based applications can be a straightforward and affordable task thanks to the increasing number of publicly accessible datasets and research publications. However, it can be extremely challenging to develop CV systems for other domain-specific applications with complex use cases, due to the tremendous amount of time and effort needed to develop custom datasets. For environments that involve human interactions (often with safety standards and compliance), creating custom datasets can also be a big challenge. In this paper, we proposed UnrealVision Dataset Generator (UVis) - a new system for generating realistic synthetic image data with automatic data labeling and annotation. Our solution aims to make use of Unreal Engine 5’s 3D rendering and animation capabilities to make it possible for researchers to create high-fidelity simulated environments with minimal efforts. We demonstrated that such a system can be used to effectively generate sufficient image data for training deep learning models with complex use cases. Nevertheless, datasets generated by UnrealVision can be modified, extended, and regenerated at any time, making them much more flexible compared to the traditional approaches. The project is available at https://github.com/lphucthinh40/unreal-vision.

Download

Share

COinS