Oceanographic Object Detection with YOLOv3

Sekaran, Sugethan (2019) Oceanographic Object Detection with YOLOv3. [Final Year Project] (Submitted)

[thumbnail of Sugethan Sekaran_20787.pdf] PDF
Sugethan Sekaran_20787.pdf
Restricted to Registered users only

Download (2MB)

Abstract

In this paper, the author uses YOLO (You Only Look Once) v3 network for oceanographic
image processing and data analysis to propose an underwater and drone object detection
system. This is mainly to address the ocean waste contamination that continuously
threatens species of marine life. The object detection model is built on top of Darknet, an
open source convolution network that makes it easy for different object detection models
to be created, trained, and executed. To be extremely fast and accurate, YOLO v3 is
graphed. Accordingly, you can easily switch between the model's speed and accuracy by
simply changing the model's size without needing any retraining. This research also
focuses on the reliability and accuracy of object detection models for testing and
documenting their output while inputting video and photographs taken with drones both
underwater and air. This is important when objects are to be accurately identified without
triggering false detection.

Item Type: Final Year Project
Subjects: Q Science > Q Science (General)
Departments / MOR / COE: Sciences and Information Technology > Computer and Information Sciences
Depositing User: Mr Ahmad Suhairi Mohamed Lazim
Date Deposited: 10 Sep 2021 08:57
Last Modified: 10 Sep 2021 08:57
URI: http://utpedia.utp.edu.my/id/eprint/20960

Actions (login required)

View Item
View Item