Logo Lanfrica
  • Accueil
  • Atlas
  • Analyses
  • Documentation
  • Sign in

© 2026 Lanfrica. Tous droits réservés. Tous les droits d'auteur des ressources affichées sur le site Web Lanfrica appartiennent aux détenteurs de droits d'auteur d'origine, sauf indication contraire explicite.

Momahmoses/crop-disease-detection-africa

Domaine:

agriculturenatural language processing

Type de record:

softwaremodel
Créateur:
Mom
Hôte:
||Mobile-ready crop disease AI providing 45-second diagnosis with treatment advice in Hausa, Yoruba and Igbo for smallholder farmers across Africa. # Crop Disease Detection for Smallholder Farmers A mobile-ready AI system that diagnoses crop diseases from phone photos in 45 seconds, with treatment advice in Hausa, Yoruba, and Igbo. Runs offline on a Raspberry Pi. ## Problem A farmer in Kaduna notices yellow spots on her maize. No agronomist is nearby. Misidentifying the disease means wrong treatment and a lost harvest, the family's income for the year. ## Quick Start ```bash pip install -r requirements.txt # Train model (uses synthetic data if PlantVillage not found) python train.py # Start inference API uvicorn src.inference.mobile_app:app --host 0.0.0.0 --port 8000 # Test diagnosis curl -X POST localhost \ -F "file=@my_crop_photo.jpg" \ -F "language=hausa" ``` ## Model Architecture - **Base**: EfficientNetV2-S pretrained on ImageNet - **Fine-tuned**: 38 disease classes across 14 crop types - **Augmentation**: blur, brightness, noise, shadow (simulates field phone cameras) - **Training**: 50 epochs, early stopping, cosine LR schedule, W&B logging - **Edge export**: ONNX → runs on Raspberry Pi 4 at <2W ## Supported Crops & Diseases ``` Apple (4 classes), Corn (4), Grape (4), Potato (3), Tomato (10), Cassava (3), Maize (4), Yam (2), Cowpea (2), Sorghum (2) ``` Total: **38 classes** ## API Response ```json { "prediction": "Tomato___Early_blight", "confidence": 94.1, "crop": "Tomato", "disease": "Early blight", "treatment": "Apply fungicide. Ensure crop rotation next season.", "treatment_localized": "Magani: Yi amfani da fungicide. Ka juyar da noman ƙasa." } ``` ## Performance - Accuracy: ~94% (with PlantVillage + augmentation) - Inference time: 45ms on GPU, 800ms on Raspberry Pi 4 - Low-confidence flag: returns "UNCERTAIN" when confidence < 60% ## Dataset Uses PlantVillage Dataset (87,000 images). Synthetic data generated automatically if dataset not present. ## Real Impact - Instant diagnosis replaces 30-minute manual inspection - Farmers in 3 languages: English, Hausa, Yoruba, …

Visit

github.com

Tasks

image classificationcomputer vision

Languages

GbagyiHausaIgboYoruba

Tags

africaagriculturecomputer-visionedge-aipythonpytorchraspberry-pi