Real-time nondestructive fish behavior detecting in mixed polyculture system using deep-learning and low-cost devices

作者：

Highlights：

• Existing improved YOLOv3-Lite provides a good balance between the accuracy and speed.

• An image enhancement phase is added to adjust for limited light transmission in water.

• Data augmentation enhances data information content and avoids over-fitting.

• An experimental system is designed, and its cost is analyzed.

摘要

•Existing improved YOLOv3-Lite provides a good balance between the accuracy and speed.•An image enhancement phase is added to adjust for limited light transmission in water.•Data augmentation enhances data information content and avoids over-fitting.•An experimental system is designed, and its cost is analyzed.

论文关键词：Fish behavior detecting,YOLO network,Low-cost imaging system,Smart fish farming,Mixed polyculture system

论文评审过程：Received 27 November 2020, Revised 21 January 2021, Accepted 14 April 2021, Available online 20 April 2021, Version of Record 30 April 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.115051