Aquaculture Europe 2025

September 22 - 25, 2025

Valencia, Spain

Add To Calendar 25/09/2025 09:45:0025/09/2025 10:00:00Europe/ViennaAquaculture Europe 2025ENHANCED FISH WEIGHT ESTIMATION THROUGH YOLO AND ELECTRODE DATA INTEGRATIONGoleta, Hotel - Floor 14The European Aquaculture Societywebmaster@aquaeas.orgfalseDD/MM/YYYYaaVZHLXMfzTRLzDrHmAi181982

ENHANCED FISH WEIGHT ESTIMATION THROUGH YOLO AND ELECTRODE DATA INTEGRATION

Jamal, A.*, Cohen, L., Hadar, O.

 Institute of Agricultural and Biosystems Engineering, Agricultural Research Organization – Volcani Institute , Rishon LeZion 7505101, Israel.

alaa@volcani.agri.gov.il



Introduction

Accurate fish weight estimation is critical to efficient aquaculture management, impacting feeding optimization, health monitoring, and harvest scheduling. Non-intrusive methods have gained significant attention for their ability to streamline this process without inducing stress in fish populations (Li et al., 2020). Among these methods, the You Only Look Once (YOLO) object detection framework has emerged as one of the most widely used approaches for non-intrusive fish weight estimation due to its high accuracy in detecting and analyzing fish morphology from images. However, YOLO’s performance can be limited when dealing with fish sizes and conditions not represented in its training data, particularly in extrapolation scenarios (Tengtrairat et al., 2022).

Method

 To address these limitations, we propose an enhanced method integrating electrode-based voltage measurements with YOLO-driven image processing, leveraging machine learning (ML) regression models for fish weight estimation. While electrode measurements have traditionally been utilized for applications such as fish counting, their integration with visual data for weight estimation remains underexplored (Sheppard & Bednarski 2015). By combining these two complementary data sources, our approach aims to improve predictive accuracy, especially in challenging conditions where image-only models struggle, with affordable component prices.

Our model was evaluated under two distinct conditions: interpolation, where YOLO performed well due to familiar weight ranges in the training data, and extrapolation, where YOLO’s detection accuracy was reduced due to unfamiliar fish sizes. The integrated model consistently outperformed image-only models in both case studies, with statistical significance. This outcome demonstrates the robustness of the combined approach, as electrode measurements provide additional physiological insights that compensate for the limitations of image-based models.

Results

 The results underscore the potential of integrating multi-modal data sources to enhance non-intrusive fish weight estimation, offering cheap and a more resilient and accurate solution for aquaculture operations.

References

  1.  Li, D., Hao , Y., & Duan, Y. (2020). Nonintrusive methods for biomass estimation in aquaculture with emphasis on fish: a review. Reviews in Aquaculture, 12(3), 1390-1411.
  2.  Sheppard, J. J., & Bednarski, M. S. (2015). Utility of single-channel electronic resistivity counters for monitoring river herring populations. North American Journal of Fisheries Management, 35(6), 1144-1151.
  3. Tengtrairat , N., Woo, W. L., Parathai , P., Rinchumphu , D., & Chaichana, C. (2022). Non-intrusive fish weight estimation in turbid water using deep learning and regression models. Sensors, 22 (14), 5161.