In Sweden, the tradition of wild berry picking is underpinned by its economic and cultural significance, particularly in providing essential nutrients in regions where fresh produce is scarce. However, the practice faces sustainability challenges, being heavily reliant on labour-intensive methods that entail significant carbon emissions due to the annual transport of around 10,000 pickers from Thailand, which generates at least 4,950 kg of CO2 per single flight. This reliance on foreign labour not only contributes to environmental degradation but also presents issues in light of restrictive immigration policies similar to those recently implemented in Finland.

Objective

The "BerryAI" project proposes creating a robust machine learning framework for the detection and classification of key wild berry species - bilberries, lingonberries, cloudberries, and crowberries - leveraging state-of-the-art convolutional neural networks (CNNs). The project will generate a comprehensive dataset to train these models, develop camera sensor-based perception systems to identify optimal setups and integrate these technologies into a prototype for field validation.

This sustainable approach aims to reduce the carbon footprint associated with berry picking, decrease dependency on seasonal labour, and enhance the yield detection of wild berries. By aligning with the needs of local SMEs and the broader agricultural sector, BerryAI seeks to ensure long-term sustainability and efficiency, contributing significantly to the local economy and setting a precedent for global agricultural practices.