Early yield prediction in different grapevine varieties using computer vision and machine learning
Problem
Predicting grapevine yield early in the season across different varieties to support wine industry yield assessment.
Why it matters
Yield assessment is highly relevant to the wine industry's planning and economics.
What the AI did
A SegNet architecture detects visible berries and canopy features from vineyard images, and support vector regression models predict actual berry counts and yield.
Results
Berry detection F1-score 0.72; R² above 0.92 for visible berries; RMSE 195 berries, NRMSE 23.83%, R² 0.79 for actual berries
Impact
Enables early-season yield estimation across six grapevine varieties in a commercial vineyard.
Record
| Venue | Precision Agriculture |
|---|---|
| Published | 2022 |
| Work type | article |
| Authors | Fernando Palacios, María P. Diago, Pedro Melo‐Pinto, Javier Tardáguila |
| Citations | 55 (OpenAlex, at import time) |
| References | 54 |
| Model or method | SegNet and support vector regression (SVR) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1007/s11119-022-09950-y |
| Open access | hybrid |
| Keywords | grapevine yield prediction, computer vision, machine learning, SegNet, support vector regression, precision agriculture |
Importance rating
real but narrower advance
Yield assessment is stated as highly relevant to the wine industry's planning and economics; demonstrated in one commercial vineyard across six varieties, a real but sector-specific advance.