Treffer: AgriSegment: A Deep Learning–Based Multi-Modal Plant Segmentation Suite for Agricultural Research

AgriSegment is a deep learning–based, multi-modal plant segmentation suite designed to bridge agricultural research and computer vision. Built with FastAPI and PyTorch, it provides web-accessible tools that make advanced segmentation workflows available to agronomists and researchers without programming expertise. The suite integrates state-of-the-art models: SegFormer (semantic segmentation), SAM – Segment Anything (interactive refinement), and Mask2Former (panoptic analysis). Four complementary modules support different research workflows: Hybrid (SegFormer + SAM): high-accuracy segmentation with feedback refinement Interactive (SAM): real-time, point-based segmentation for single images Semantic (SegFormer): automated batch processing for large datasets Panoptic (Mask2Former): instance, semantic, and panoptic segmentation with detailed statistics AgriSegment enables reproducible pipelines for precision agriculture, crop phenotyping, and computational agro-ecology, while also serving as a testbed for computer vision and AI researchers working on segmentation methods. This release corresponds to v1.0.0, archived for long-term preservation. Future versions will be linked to the accompanying ICROPM 2026 conference paper and a forthcoming journal publication. FundingThis work was supported by the Italian Space Agency (ASI) under the project “An Open, Efficient, and Customizable Pipeline for the Automated Processing of Remote Sensed Data for Computational Agro-Ecology”, and by Regione del Veneto under PR FESR 2021–2027, Action 1.1.1 in the context of the project “AGRIFUTURE: Il futuro della sostenibilità per le sfide competitive delle aziende agroalimentari venete” (ID 24279_001587_04499230235). The funders had no role in study design, code development, or manuscript preparation. LicenseLicensed under the MIT License. See the LICENSE file for details.