Object detection using machine learning techniques

• Main Author: Hu, Xiaoxiang
• Other Authors: Huang Guangbin
• Format: Final Year Project
• Language: English
• Published: 2018
• Subjects: DRNTU::Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
• Online Access: http://hdl.handle.net/10356/74913
• Institution: Nanyang Technological University

The application of machine learning techniques in object detection area has been improved dramatically for recent 5 years. As a result, different research groups around the world proposed numbers of state-of-the-art object detection models. However, with the development of autonomous driving technology, new requirements for object detection task are raised as well. This project mainly focuses on the analysis and optimization on state-of-the-art deep learning models. A brief study on basic deep learning knowledge, especially the mathematic model and the functionalities of components in this area, are performed. This project conducts a systematically analysis on YOLO and YOLOv2 model. Some minor adjustments on original YOLOv2 model are performed to fit our customized local dataset. In addition to these, preliminary analysis on a novel deep learning model named RCN is included in this project. Despite using feedforward network, RCN creates lateral connections among latent variables in internal layers, to mimic the functionalities of human brain. With the help of RCN, the object detection model is able to recognize the shape and appearance of an object. However, the experiment of RCN only limits to text-based CAPTCHA currently, and some further studies on general object detection using RCN are still required