Welcome to Plant Pathology AI

Diagnosing crop health with computer vision to ensure food security for the future.

PP

Plant Pathology — AI Plant Disease Detection

End-to-end pipeline for detecting plant diseases from leaf images and suggesting targeted remedies.

Image classification Transfer learning Edge deployable Kiswahili-ready UI
Plant pathology demo image
Model: EfficientNet-B3 (finetuned)
Input: 512x512 RGB
Note: demo here runs a lightweight client-side mock for fast preview. Full inference happens on server or edge device in production.

Results & Metrics

mAP (val)
0.92 overall
Top-1 Acc
0.94
Inference
~45 ms (edge optimized)
Model size
~18 MB (quantized)

How it works (brief)

  1. Collect labeled leaf images; ensure diversity across lighting and cultivars.
  2. Augment with rotations, color jitter, gaussian blur, and random crops.
  3. Train using transfer learning and appropriate loss for imbalance.
  4. Export model (TorchScript/ONNX), optimize with pruning/quantization for deployment.
  5. Provide UI for farmers with local-language support, step-by-step remedies, and record-keeping.