Deep Learning
AI-Powered Fertilizer Deficiency Detection
A full-stack intelligent agriculture system that lets farmers and agronomists upload a plant/soil image or input crop data and instantly detect nutrient deficiencies — Nitrogen, Phosphorus, Potassium, Magnesium, Iron and Calcium — then receive AI-driven fertilizer recommendations. Three architectures were benchmarked for the detection engine: a custom CNN, ResNet-50, and EfficientNet-B3, which won out with 96.3% accuracy. Shipped as a role-based (farmer/agronomist/admin) Next.js + FastAPI web app, containerized with Docker Compose.
Model Accuracy
96.3%
Model Type
EfficientNet-B3 (CNN)
Category
Deep Learning
Tech Stack
8 Tools
Key Features
- ✓Deep learning multi-class deficiency detection from leaf images
- ✓Soil-data analyzer with NPK & pH-based quality scoring
- ✓AI chat assistant for natural-language crop-health questions
- ✓Fertilizer recommendation engine tailored to crop type and detected deficiency
- ✓One-click auto-generated PDF diagnosis reports
- ✓Role-based access (farmer / agronomist / admin) with crop-history tracking
- ✓Dockerized production deployment behind Nginx
Technologies
TensorFlowOpenCVFastAPINext.jsTypeScriptMySQLDockerNginx
Project Screenshots
Performance Metrics
Best Accuracy
96.3%
Precision
96.0%
Recall
96.5%
F1-Score
96.2%