Project 11 · Applied Vision
BIO
PlantPathologist
Plant Disease Detection
01 Abstract
Plant Disease Detection
A mobile-friendly disease detector that identifies 18 diseases across five crop species from leaf photos and pairs each diagnosis with pathogen-appropriate treatment guidance.
02 Highlights
Methods & contributions
- 23-class taxonomy — 18 diseases plus 5 healthy classes across tomato, potato, corn, apple, and grape.
- Pathogen-aware advice — correctly types oomycete (late blight), bacterial, and viral diseases; recommends vector control rather than fungicide for viral cases.
- Transfer learning — EfficientNet-B0 with temperature-scaled confidence calibration.
- Field-ready — camera-enabled web interface with an accurate Latin-name disease knowledge base.
03 Architecture
Data flow
leaf photo → EfficientNet-B0 → 23-class softmax
→ disease database → pathogen type + treatment
04 Stack