Berl’Eyes: a tool for quickly detecting and quantifying rapeseed flea beetle larvae using a smartphone

A Key Challenge for Rapeseed Protection

Rapeseed is one of the field crops that uses the most plant protection products. Among its main pests is the large flea beetle. Adult insects feed on the leaves, while the larvae develop inside the stems, weakening the plants and potentially leading to significant yield losses. In the face of this risk, prevention relies largely on the ability to detect the pest early and assess the level of infestation.

A Cumbersome Traditional Method

Counting flea beetle larvae traditionally relies on the Berlese method. It involves placing rapeseed stems over a container to allow for the passive extraction of larvae after drying. The larvae fall to the bottom of the container and are then counted manually. This final counting step is particularly time-consuming and tedious.

The Role of Artificial Intelligence

Berl’Eyes simplifies this process. A simple photo of the container after extraction is all it takes to determine the number of larvae in just a few seconds.

The app uses an automatic larva identification model, trained using approximately 500 annotated images of various collection basins containing different quantities of larvae. Efforts were also made to standardize the photos to ensure reliable results.

Automated Counting for Agroecology

Berl’Eyes is a concrete example of how digital technology can aid in the phenotyping of pests and contribute to practical solutions for the agroecological transition.

By automating and accelerating the counting of larvae, the tool facilitates access to data on the flea beetle and supports the development of new strategies to combat this pest: variety selection, changes in farming practices, and optimization of plant protection products. Berl’Eyes also provides a key indicator for decision-making in flea beetle management, contributing to more precise intervention planning, with better-targeted and potentially less frequent treatments.

The tool is thus fully aligned with the objectives of the agroecological transition, contributing to more sustainable crop management and meeting the expectations of the oilseed and protein crop sector.

Future prospects include distinguishing between larval development stages, regulatory recognition of the method, mapping observations to inform epidemiological surveillance bulletins, and the identification of other insect pest species directly in the field, including the large flea beetle.

A development driven by several research projects

Berl’Eyes was initiated by the RESALT project (led by INRAE, Innolea, Terres Inovia, and ten plant breeders), as part of the plan to phase out Phosmet, with funding from the Ministry of Agriculture, Agri-Food, and Food Sovereignty via the CASDAR fund, as well as support from the vegetable oils and proteins sector via Sofiprotéol’s FASO fund. The RESALT project enabled the development of an initial version of the model and application, validated the proof of concept, and carried out the first data acquisitions.

Subsequently, the AgroEcoPhen project under the PEPR Agroecology and Digital initiative contributed to its evolution by leveraging the capabilities of the Phenome-Emphasis phenotyping infrastructure and by funding developments to update and disseminate the model. This support has notably enabled improvements to the model’s training methods, the enrichment of the datasets, and the updating of the web application’s IT infrastructure.

Access to the tool and official launch

The tool is freely available online: https://berleyes.terresinovia.fr/

The official launch will take place on May 21, 2026, during an informational webinar organized by Terres Inovia, from 1:30 PM to 2:00 PM. To register: https://www.terresinovia.fr/fr/evenements/berleyes-un-outil-pour-mieux-detecter-les-larves-daltises-grace-lintelligence