Year: 2026
Type: report
Abstract: Many fish populations of interest for fisheries are still in an unknown state due to a lack of data. Samplings in fishmarkets and during scientific campaign produce many usefull data but their costs are considerable. The sampling can then target few species during a restricted time only. In the same time, the French fishmarkets equipped themselves with photographic equipments, meaning that most landed fish boxes (sorted by specie and commercial category) are pictured for online auctions. The method developed in the FishSizeMe II project offers a mean to use those pictures from fishmarkets to estimate, using deep learning models, the lengths frequencies of the landed fishes. There are two steps in the estimation process of these frequencies. First, the localisation and the size of the fishes are predicted by the deep learning models for most of those that appears in the pictures. However not all fishes are predicted. An extrapolation step based on the landed weight and length-weight relationships is then needed to estimate the landed length frequencies. The endgame of this project is having lengths frequencies usable in length-based methods to estimate the state of fish populations.
Author: Lemaitre, Aloys
Language: English
Subfield: Aquatic Science
Domain: Life Sciences
SDG: Life below water
Open Access status: green
OpenAlex ID: W7221733988