PRACTICAL WISION – Automatically identifying and mapping different weed species through the practical application of AI-based image analysis models

Weeds are the pest with the highest crop yield loss potential and sustainable weed management is critical for agriculture to feed a growing population while maintaining the ecosystems and biodiversity that support humanity.

To design more sustainable weed management strategies, weed researchers need more powerful tools than traditional weed sampling, which produces high-quality but subjective data that is limited in scope and in spatial information.

Artificial intelligence (AI)-based machine vision techniques can be used to automatically identify and classify weed species and can potentially be used to provide quick and objective mapping of weed populations over huge areas.

While this use of AI-based machine vision has great theoretical potential it is currently not used in practice in Swedish agriculture.

To address this, PRACTICAL WISION will 1) evaluate how well AI-based models can identify, classify and map different weed species in Swedish field trials, 2) build up an iterative process to collect high-quality and annotated weed images in Swedish field trials (to be published in open-access) – which will lead to improved weed recognition over time, and 3) further develop the AI models to become more suitable for use in field trials.

PRACTICAL WISION represent most of the organizations that carry out field trials in Sweden: RISE Research Institutes of Sweden, the Swedish University of Agricultural Sciences, Hushållningssällskapet Skåne and Nordic Beet Research.