Pré-Publication, Document De Travail Année : 2025

Automated Counting of Fish in moving Diver Operated Videos (DOV) for Biodiversity Assessments

Résumé

1 - Underwater video transects are crucial to assess marine biodiversity and biomass. The counting of individual fish in these videos is labour- and time-intensive and operator-dependent. Automating this step would create non-biased biodiversity data and decouple the data collection campaign from any field constraints. 2 - To begin developing an automated process, we assessed the commonly used method for counting objects in videos and compared it to three new methods for counting fish using computer vision derived data (single frame detections) that resulte in a holistic and fully automated pipeline for fish abundance measurements. In addition to the commonly used method Nmax, we included (i) a 1D k-means clustering method, (ii) an intuitive clustering approach, NHeuristic, and (iii) a Temporal Convolutional Neural Network (TCN) counting method. We tested these methods on three Mediterranean species from different ecological niches. 3 - We first assessed the methods using manually labelled detections (groundtruth detections) and then incorporated a detector into the pipeline for a more realistic scenario. The results in these two configurations showed evidence of underestimation by Nmax. The other methods showed better overall results. The proposed NHeuristic and TCN methods are the closest to manual evaluation. With an absolute variation comparable to inter-operator variation, we demonstrated that these are reliable methods for quantifying fish counts for these three different Mediterranean species. 4 - The parameters of these automated methods could be adjusted to suit other species and then be used in monitoring programs, for example to assess biodiversity and biomass in marine protected areas over time.
Fichier principal
Vignette du fichier
fishCount_Paper_noLineNumbers.pdf (14 Mo) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04865293 , version 1 (06-01-2025)
hal-04865293 , version 2 (28-01-2025)

Identifiants

  • HAL Id : hal-04865293 , version 2

Citer

Kilian Bürgi, Rémy Sun, Charles Bouveyron, Diane Lingrand, Benoit Dérijard, et al.. Automated Counting of Fish in moving Diver Operated Videos (DOV) for Biodiversity Assessments. 2025. ⟨hal-04865293v2⟩
14 Consultations
18 Téléchargements

Partager

More