Automated Recognition of Underwater Objects using Deep Learning
Sowmya Raavi, Phani Babu Chandu, T. Sudalaimuthu
Abstract
The underwater object detection is a difficult task due to the unclear visibility and lack of objects underwater in kovalam, it was expected to have visibility up to three to five meters but it is very complicated to detect deeper/underwater objects like shoals of anchovy and silver moony, Bluefin trevally, puffer fish, box fish, etc., The existing method focuses on identifying a single object underwater and have low accuracy (78%) to detect multiple objects in a single instance. The proposed solution intends to improve the accuracy of identifying objects underwater. The proposed work can be used in submarines, military purposes, and archeology to identify objects effectively. The proposed system employs novel techniques for precise object detection and techniques like YOLO and SSD along with precise object detection methods and algorithms like Convolutional Neural Networks (CNN). These artificial intelligence algorithms allow to detect every object in an image by the area it takes up in a highlighted rectangular box, identify every object, and tag every object.