Shark-EYE: A Deep Inference Convolutional Neural Network of Shark Detection for Underwater Diving Surveillance
2021 International Conference on Computational Intelligence and Knowledge Economy (ICCIKE)
(2021), pp. 384-388
Nino E. Merencilla
a
,
Alvin Sarraga Alon
b
,
Glenn John O. Fernando
c
,
Elaine M. Cepe
d
,
Dennis C. Malunao
c
a Department of Computer Engineering, FEU Institute of Technology, Manila, Philippines
b Digital Transformation Center, STEER Hub Batangas State University, Batangas City, Philippines
c College of Computing Sciences, Ifugao State University, Ifugao, Philippines
d College of Engineering and Architecture, Bohol Island State University, Tagbilaran, Philippines
Abstract: People are anxious about the potential dangers of scuba diving and like in all sports, there are dangers involved in it. Typically, people think sharks and shark attacks are the dangers of scuba diving, as sharks are one of the ocean's biggest predators, and the great white shark, in particular, is one of the primary threats to divers. The study proposes a deep learning approach to shark detection for underwater diving surveillance. A large collection of great white sharks' datasets underwater is used by the system for training as sharks are hard to differentiate from other sharks like animals in an underwater environment. A YOLOv3 algorithm that uses convolutional neural networks for object detection, multiscale prediction, and bounding box prediction through the use of logistic regression is used in the study. And with this approach, the testing of the shark detection system generates a good result.