Related Experiment Video
Updated: Jul 10, 2026
Using an Automated 3D-tracking System to Record Individual and Shoals of Adult Zebrafish
Published on: December 5, 2013
Enhancing eco-sensing in aquatic environments: Fish jumping behavior automatic recognition using YOLOv5
Kaibang Xiao1, Ronghui Li1, Senhai Lin1
1College of Civil Engineering and Architecture, Guangxi University, Nanning 530004, PR China; Key Laboratory of Disaster Prevention and Structural Safety of the Ministry of Education, College of Civil Engineering and Architecture, Guangxi University, Nanning 530004, PR China.
This study introduces a novel fish jumping behavior (FJB) recognition model using YOLOv5, achieving over 97% accuracy. The enhanced model reliably detects fish activity above water, aiding aquatic ecology and aquaculture assessments.
Area of Science:
- Aquatic ecology
- Behavioral ichthyology
- Computer vision in environmental monitoring
Background:
- Traditional fish behavior studies are limited by underwater observation challenges.
- Supra-aquatic observation offers a clearer perspective on fish physiology and habitat.
Purpose of the Study:
- To develop and validate a fish jumping behavior (FJB) recognition model.
- To improve fish behavior monitoring in aquatic ecosystems.
- To establish a foundation for intelligent aquatic ecology assessment.
Main Methods:
- Utilized the YOLOv5 convolutional neural network for target detection.
- Trained and validated a model on 877 images of fish jumping from reservoir data.
- Developed an enhanced YOLOv5-SN model incorporating ripple variation and duration rules.
- Tested model robustness across various weather conditions (rain, cloudiness, sunshine).
Main Results:
- YOLOv5 demonstrated superior splash detection compared to YOLOv7, YOLOv8, and YOLOv9.
- Achieved >97% precision and recall with an F1 score >0.9 on the validation set after 50 epochs.
- The enhanced YOLOv5-SN model effectively mitigated noise interference.
- The model showed robustness in detecting fish jumping behavior under diverse meteorological conditions.
Conclusions:
- The YOLOv5-based model provides an accurate and robust method for supra-aquatic fish behavior recognition.
- The developed system can significantly advance intelligent perception in aquatic ecology and aquaculture.
- Recognizing fish jumping behavior offers insights into fish physiology and habitat conditions.
