Motorcycle Detection using Deep Learning Convolution Neural Network

Detecting and avoiding motorcycles on roads is important for Autonomous Vehicle (AV). This is because a majority of accidents occurring in Malaysia involve motorcycles. Detecting motorcycles is a challenging task due to its low visibility and high velocity. This research attempts to capitalize on De...

Full description

Saved in:
Bibliographic Details
Main Authors: Ismail, Fatin Natasha, Yassin, Ihsan Mohd, Ahmad, Adizul, Ali, Megat Syahirul Amin Megat, Baharom, Rahimi
Format: Conference Proceeding
Language:English
Subjects:
Convolution
Deep learning
Faster Region Convolution Neural Network
Feature extraction
motorcycle detection
Motorcycles
object recognition
Roads
Task analysis
Training
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Detecting and avoiding motorcycles on roads is important for Autonomous Vehicle (AV). This is because a majority of accidents occurring in Malaysia involve motorcycles. Detecting motorcycles is a challenging task due to its low visibility and high velocity. This research attempts to capitalize on Deep Learning Neural Network to detect motorcycles. Training involves various motorcycle models and poses with different resolutions and road conditions. The AlexNet network structure was chosen for implementation due to its proven performance in object detection tasks. Transfer learning was used to repurpose the AlexNet network for the described task. Training and classification were performed using the MATLAB Deep Learning Toolbox. Test results on our custom dataset demonstrates the effectiveness of the approach for the task.
ISSN:2470-640X
DOI:10.1109/ICSET51301.2020.9265361