About the project
Farmers often find a leaf disease late, or spray the wrong product in the wrong quantity. Plant leaf disease detection Android project LeafCure solves this with a deep learning model that runs inside the phone. A MobileNetV2 CNN trained on the PlantVillage dataset (54,303 images, 38 classes, 14 crops) identifies the disease in seconds, even without internet. Image segmentation then measures how much of the leaf is affected, and a dosage calculator tells the farmer exactly how much product and water to use.
How it works
- Farmer selects the crop and takes a leaf photo (camera or gallery)
- The TFLite CNN shows the top 3 diseases with confidence
- Leaf segmentation gives severity: Mild, Moderate or Severe
- The app shows organic and chemical remedies with dose, water, spray schedule and cost for the plot
- The scan is saved with photo and GPS; low-confidence scans go to the agriculture officer
- The server finds disease clusters by distance and days and creates outbreak alerts
- The officer views the disease map and alerts all farmers nearby
Users
- Farmer: scan leaves, remedies and dose, history with PDF report, disease library, ask officer, quiz, progress
- Agriculture Officer: verify or correct scans, disease map, outbreak alerts, advisories, answer questions
- Admin: crops, diseases, remedies, users, scans, outbreaks, quiz, notices, reports and settings
Highlights
- 3 roles, 24 Android screens, 13 admin pages, 30 PHP APIs, 13 MySQL tables
- On-device CNN (MobileNetV2 + TensorFlow Lite) with offline quick-check fallback
- Severity estimation by HSV image segmentation, plus a farm-size dosage calculator
- Outbreak detection with the haversine distance rule and an officer disease map
- Validation in both app and PHP, hashed passwords, prepared statements, CSRF-protected admin
- Demo data with ready logins for farmer, officer and admin
What you get
- Complete Android Studio source code (Java) and PHP admin panel
- MySQL database file with demo data
- Google Colab notebook and Python script to train the CNN and export TFLite
- Project report (Word, 50+ pages) with DFD, ER and UML diagrams
- 2-page project abstract, project explanation with viva questions
- Setup guide and installation help
Who can use this project
Final year BE / B.Tech, ME / M.Tech, MCA, BCA and Diploma students of Computer, IT and ENTC departments looking for a deep learning Android project, a machine learning project with source code, an agriculture app project or a TensorFlow Lite image classification project with documentation.
Frequently asked questions
Does the disease detection need internet? No. The CNN runs on the phone. Internet or Wi-Fi is needed only to save scans and use the officer features.
Is the trained model included? A ready Colab notebook trains it in about 30 minutes on the free GPU. Until then, the app runs with the offline analyzer.
Which crops are supported? 14 crops of PlantVillage, including tomato, potato, grape, maize, apple, capsicum and orange (38 classes).
Can I add new diseases? Yes. Add them in the admin panel with the same model label as in labels.txt.
