Comparative evaluation of two automated fish counting software tools using acoustic camera data
Abstract
Acoustic cameras are powerful tools for studying fish behaviour, including spawning phenology, even in turbid conditions where optical systems fail. Monitoring whitefish (Coregonus sp.) spawning in lakes poses specific challenges: fish may re-enter the camera's field multiple times, producing indices of activity rather than true counts. Manual data processing, although reliable, is labour-intensive, while existing semi-automatic software can be complex and costly. We aimed to test the performance of the open-source software, FishTracker (FishT) for monitoring whitefish spawning phenology and compare its outputs with manual counts and those of the widely used Sonar5-Pro (S5P). Acoustic camera data were collected during two spawning seasons: one under normal weather conditions (2020) and one with strong wind weather conditions (2023). Manual counts by two operators were used as a reference. The same datasets were processed with S5P and FishT. Lin's concordance correlation coefficient was used to assess agreement between methods, and qualitative assessments of software usability and stability were conducted. Our results indicated that FishT and S5P showed strong agreements with manual counts. Wind-induced wave action reduced both software's detection performance, though FishTracker calibration under favourable conditions limited overcounting despite suboptimal phenological trend capture. FishT processed large datasets efficiently and was user-friendly but tended to over-count under degraded conditions. FishT demonstrates strong potential as an accessible, low-cost alternative for analysing acoustic camera data. Limitations remain in adverse weather, highlighting the need for further development, particularly in detection algorithms, to improve reliability and broaden its use in aquatic monitoring.
Related
- Brief communication: Landsifier 2 – Python-based Software for Automated Landslide Type Classification
- Automated photogrammetric close-range imaging of small invertebrates using acoustic levitation
- Comparative analysis of fish sounds recorded during an early acoustic experiment in a Mediterranean seagrass meadow
- Evaluation of Distributed Acoustic Sensing (DAS) Systems with Polarization Diversity for Railway Infrastructure: Effects of Fiber Type and Installation Conditions on System Performance
- AutoCF: An Automated LLM-Assisted Ecosystem for Compound Flood Simulation, Evaluation, and Impact Attribution
BibTeX
@article{godeaux2026comparative,
title = {{Comparative evaluation of two automated fish counting software tools using acoustic camera data}},
author = {Quentin Godeaux and Hervé Rogissart and Clément Rautureau and François Martignac and Franck Cattanéo and Jean Guillard},
journal = {Ecological Informatics},
year = {2026},
month = sep,
doi = {10.1016/j.ecoinf.2026.103904},
url = {https://hal.science/hal-05734532v1},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Comparative evaluation of two automated fish counting software tools using acoustic camera data
AU - Quentin Godeaux
AU - Hervé Rogissart
AU - Clément Rautureau
AU - François Martignac
AU - Franck Cattanéo
AU - Jean Guillard
PY - 2026
DA - 2026/09/01
JO - Ecological Informatics
DO - 10.1016/j.ecoinf.2026.103904
UR - https://hal.science/hal-05734532v1
AB - Acoustic cameras are powerful tools for studying fish behaviour, including spawning phenology, even in turbid conditions where optical systems fail. Monitoring whitefish (Coregonus sp.) spawning in lakes poses specific challenges: fish may re-enter the camera's field multiple times, producing indices of activity rather than true counts. Manual data processing, although reliable, is labour-intensive, while existing semi-automatic software can be complex and costly. We aimed to test the performance of the open-source software, FishTracker (FishT) for monitoring whitefish spawning phenology and compare its outputs with manual counts and those of the widely used Sonar5-Pro (S5P). Acoustic camera data were collected during two spawning seasons: one under normal weather conditions (2020) and one with strong wind weather conditions (2023). Manual counts by two operators were used as a reference. The same datasets were processed with S5P and FishT. Lin's concordance correlation coefficient was used to assess agreement between methods, and qualitative assessments of software usability and stability were conducted. Our results indicated that FishT and S5P showed strong agreements with manual counts. Wind-induced wave action reduced both software's detection performance, though FishTracker calibration under favourable conditions limited overcounting despite suboptimal phenological trend capture. FishT processed large datasets efficiently and was user-friendly but tended to over-count under degraded conditions. FishT demonstrates strong potential as an accessible, low-cost alternative for analysing acoustic camera data. Limitations remain in adverse weather, highlighting the need for further development, particularly in detection algorithms, to improve reliability and broaden its use in aquatic monitoring.
SN - 1574-9541
ER -