Design

a comprehensive pipeline generating digital maps that provide more comprehensive information about the vineyard's condition.

Assess

nutritional deficiencies in vines.

Integrate

remote sensing and ground data in real-life scenarios in vineyards.

DiVine

Vele Sense as a start-up from Serbia, leverages AI and multispectral imaging for early grapevine stress detection, identifying diseases, water shortages, and nutrient deficiencies. The goal is to equip vineyard owners with a monitoring tool for timely interventions and yield protection.

waves

DJI Mavic 3 Multispectral

RTKmobile station

GPUs for model(s)

RGB and multispectral drone images
Commercial software for creation of orthomosaics (e.g., Pix4D)
Disease detection method (VelesSense proprietary software)
Python programming language and Python scripts: VI calculation, ROI extraction, Vines extraction, Sampling zones calculation, JSON2JPG (open access), Lab for sample analysis

Outcomes

Creation of ‘readable’ ground truth maps, from bounding boxes to full marks.

Veles Sense user-friendly interface (available with an invitation).

Challenges Faced

Maps georeferencing needed to be done manually

Automatic vine detection (fully relying on supervised learning) requires a big dataset for training that was not feasible to achieve during the project duration

Lab analysis from two different labs showed significantly different results.

Impact (Socio-economic & Environmental)

Socio-economic

Alignment with Smart Specialization Strategy of the Republic

of Serbia 2020-2027

Enhanced economic resilience, promoted sustainable practices, and supported market growth for vineyards.

Environmental

Alignment with EU strategies such as ‘’From Farm to Fork Strategy’’, part of the European Green Deal.