Plant Doc: On-Device Plant Disease Detection
Built a mobile plant disease assistant that lets users capture or upload a plant image, runs MobileNetV2-based TensorFlow Lite inference locally, and returns a disease label, confidence score, Turkish explanation, and care recommendation.
Classes
0
Plant health and disease labels
Input
0x224
Model image tensor size
Inference
On-device
No backend server required
Checks
0
Analyze, tests, Android APK build
Project Gallery
1/4
Problem
Plant disease inspection usually requires expert knowledge, and cloud-based AI workflows add latency, connectivity, and privacy constraints for simple field use cases.
Challenge
The main challenge was packaging an ML model inside a Flutter app, preparing camera/gallery images into the expected tensor format, and keeping the app buildable across modern Flutter and Android tooling.
Architecture
How the pieces fit together.
The Flutter UI collects an image through camera or gallery, passes it to a ModelService layer, resizes it to 224x224 RGB, normalizes pixel values, runs TensorFlow Lite inference, maps the output to labels, and converts the prediction into user-facing disease guidance.