Basic Study of Deep Learning Based Efficient Hermit Crabs Detection from Drone-Captured Images

Abstract

The challenges arising from water clarity, depth, and other factors intensify the difficulties in surveying underwater hermit crabs, exacerbated by a notable shortage of practical field surveys. This study introduces a novel approach utilizing consumer-grade Unmanned Aerial Vehicles (UAVs) and deep learning to investigate underwater hermit crabs. We applied diverse super-resolution algorithms, employing distinct design strategies for image enhancement. Furthermore, we utilized the proposed object detection model developed from YOLOv8, achieving a mean average precision (mAP) of 0.722, surpassing other state-of-the-art object detection algorithms. Applying UAVs and super-resolution technology has significantly progressed underwater hermit crab detection, providing practical solutions for aquatic ecological monitoring, and enabling precise benthos detection.