HERSS-MAIZE 2025; High-rEsolution dRone dataset for Site-Specific weed wanagement in Maize (San Piero - Pisa, Tuscany, Italy)

Weed management (WM) is recognised as one of the major challenges in modern agriculture, particularly within the framework of the Farm to Fork strategy, which aims to reduce pesticide use by 50% by 2030 while maintaining high crop productivity. Precision agriculture techniques, and especially Site-Specific Weed Management (SSWM), have emerged as promising approaches to achieve these goals by optimising post-emergence herbicide application according to the spatial distribution of weeds. The SSWM process is characterised by three fundamental phases: weed detection (WD), potential damage (PD) estimation, and precision herbicide application (PHA), all supported by advanced sensing, imaging, and artificial intelligence (AI) methods.

The first HERSS-MAIZE dataset was acquired in 2024 over a maize experimental field measuring approximately 0.8 ha, located in San Piero (Pisa, Tuscany, Italy), during the crop’s critical period of susceptibility to weed competition. To ensure temporal continuity and enable multi-year analyses of weed dynamics and crop–weed interactions, a subsequent acquisition was conducted in 2025 over the same experimental site, following a comparable acquisition protocol.

The primary objective of the HERSS-MAIZE dataset is to provide a high-resolution, multi-temporal UAV-based resource for developing, validating, and benchmarking algorithms and methodologies for Weed Detection and Site-Specific Weed Management in maize. The incorporation of data from both professional and consumer-grade UAVs within the dataset facilitates research into scalable, cost-effective precision agriculture solutions aimed at reducing herbicide use and enhancing environmental sustainability. Ultimately, HERSS-MAIZE supports advancements in AI-driven weed mapping, adaptive management zone delineation, and precision spraying strategies aligned with sustainable agricultural practices.