Rahma-Shaban/Zar3y-Explainable-Crop-Disease-Detection-for-Smallholder-Farmers
Domaine:
agricultureType de record:
modelsoftware
Créateur:
Rah
Hôte:
Zar3y ("My Crop") is an end-to-end AI solution designed to empower Egyptian smallholder farmers in the Nile Delta and Upper Egypt. By leveraging computer vision, the system provides immediate diagnosis of crop diseases—such as those affecting tomatoes, potatoes, peppers, and corn—mitigating the heavy losses caused by delayed expert intervention.
# 🌱 Zar3y — Crop Disease Detection from Phone Photos
> **"Hassan, a smallholder farmer in Beheira, sees yellow spots on his tomato leaves. He takes a photo with his phone."**
Zar3y ("my crop") identifies crop leaf diseases from a single photo — returning a **plain-language diagnosis**, **confidence score**, and **Grad-CAM overlay**.
---
## Team (5 Members)
| Role | Member | Responsibilities |
|------|--------|-----------------|
| Data Engineer | Nourhan | Data prep, augmentation, OOD test set |
| ML Engineer (Train) | shahd | Transfer learning, training pipeline |
| ML Engineer (Eval) | menna | Evaluation, metrics, benchmarks |
| Backend Developer & team leader | Rahma | FastAPI, TFLite inference, quantization |
| Frontend Developer | tasneem | Streamlit demo, Grad-CAM UI, demo recording |
## Dataset
- **Source**: PlantVillage on Kaggle
- **Subset**: 10 locked classes (Tomato, Potato, Pepper, Corn)
- **Split**: Stratified 70/15/15, **seed=42**
### 10 Locked Classes
| # | Class | Crop |
|---|-------|------|
| 0 | Tomato___healthy | Tomato |
| 1 | Tomato___Early_blight | Tomato |
| 2 | Tomato___Late_blight | Tomato |
| 3 | Tomato___Leaf_Mold | Tomato |
| 4 | Potato___healthy | Potato |
| 5 | Potato___Early_blight | Potato |
| 6 | Potato___Late_blight | Potato |
| 7 | Pepper_bell___healthy | Pepper |
| 8 | Pepper_bell___Bacterial_spot | Pepper |
| 9 | Corn___Common_rust | Corn |
### Per-Class Distribution
| Class | Train | Validation | Test |
|---|---|---|---|
| Corn_(maize)___Common_rust_ | 834 | 179 | 179 |
| Pepper,_bell___Bacterial_spot | 697 | 150 | 150 |
| Pepper,_bell___healthy | 1034 | 222 | 222 |
| Potato___Early_blight | 700 | 150 | 150 |
| Potato___Late_blight | 700 | 150 | 150 |
| Potato___healthy | 106 | 23 | 23 |
| Tomato___Early_blight | 700 | 150 | 150 |
| Tomato___Late_blight | 1336 | 286 | 287 |
| Tomato___Leaf_Mold | 666 | 143 | 143 |
| Tomato___healthy | 1113 | 239 | 239 |
> Note: `Potato___healthy` contains fewer than 500 images. Class …
Similaires
AI-Driven Mobile Crop Disease Detection for Smallholder Farmers: A Literature Review and Implications for ZimbabweAI-Based Cassava Leaf Disease Detection to Support Smallholder Farmers in GhanaDataset for Crop Pest and Disease Detectionmichaelnkema1/crop-disease-detectionCrop Biotechnology and Smallholder Farmers in AfricaSghaier04Mouhanned/ghana-crop-disease-detection