Basic Study of Deep Learning Based Efficient Hermit Crabs Detection from Drone-Captured Images
Fan Zhao, Dianhan Xi, Yijia Chen, Bangzhang Ma, Yongying Liu, Jiaqi Wang, Katsunori Mizuno
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.