Abstract
This applied research project develops a low-cost, mobile-based computer vision system that enables smallholder farmers to detect crop diseases early, reducing yield losses by up to 40% and minimizing pesticide overuse.
Research Methodology
The research employed a mixed-methods approach combining field trials across 12 districts, deep learning model development using transfer learning on 50,000+ annotated crop images, and participatory design sessions with 200+ farmers.
Key Findings
- Achieved 94.2% disease classification accuracy on field-tested images
- Reduced average pesticide application by 35% in pilot farms
- Demonstrated ROI of 3.2x for farmers within one growing season
Publications
- Sharma, A. et al. (2025). Mobile CV for Precision Agriculture. Journal of Applied AI.