About the project
Most city waste in India reaches landfills mixed because people are not sure which bin an item belongs to, and nobody checks or rewards good segregation. Smart waste classification Android project EcoSort solves this with the phone camera: a MobileNetV2 CNN running on the phone (TensorFlow Lite) identifies the waste type in about a second, shows the correct bin and how to dispose of it, and the same model checks segregation when the waste is collected.
How it works
- The citizen takes a photo (camera or gallery) and the TFLite model classifies it into plastic, paper, metal, glass, organic, e-waste or reject
- The result screen shows confidence, the bin in its colour, disposal steps and the top 3 possibilities
- The scan is saved and earns green points (minimum confidence and a daily limit stop misuse); a streak counts daily scanning
- Citizens book a door pickup; the collector whose weekly route covers the ward is assigned automatically
- At the door the collector photographs the bags, enters the citizen's OTP and weight; the segregation score gives bonus or penalty points
- Points buy rewards, and a ward leaderboard, segregation quiz and monthly eco report (PDF) keep citizens engaged
- Wrong results can be reported; accepted corrections become new training photos
Users
- Citizen: scan waste, disposal guide, pickups with OTP and rating, points, rewards, quiz, eco report
- Collector: today's route, assigned pickups, bag segregation check, collect / not available, schedule, history chart
- Admin (web): dashboard, approvals, wards, categories, quiz, rewards, pickup assignment, routes, scan review, reports
Highlights
- 3 roles, 22 Android screens, 13 admin pages, 31 PHP APIs, 13 MySQL tables
- On-device CNN (MobileNetV2 transfer learning, 7 classes) with Colab training notebook and offline Quick mode fallback
- Same model used twice: item classification and pickup segregation score
- Validations in app and server: email, 10-digit mobile, password, pincode, OTP, weight range, date picker, photo type and size, duplicates
- Prepared statements, password hashing, CSRF token in admin forms, double-submit protection
- Demo data with Pune wards, citizens, collectors, scans and pickups, plus ready demo logins
What you get
- Complete Android Studio source code (Java) and PHP source code
- MySQL database file with sample data
- Model training notebook (Google Colab) and training script
- Project report (55 pages, Word) with all UML and DFD diagrams
- Project abstract, project explanation with 21 viva questions, setup guide
- Setup help to run the project on your laptop
Who can use this project
Final year BE / B.Tech (Computer, IT, E&TC), MCA, BCA and Diploma students looking for an Android project with source code and documentation, especially in machine learning, deep learning, image classification, smart city or Swachh Bharat themes.
Frequently asked questions
Does the waste classification need internet? No. The TFLite model runs on the phone; internet is needed only to save scans and pickups to the server.
How is the model trained? Open the given notebook in Google Colab and click Run all; it downloads free Kaggle garbage datasets, trains MobileNetV2 and gives the .tflite file for the app.
Can I run it without a real phone? Yes, the Android emulator works with the default server address 10.0.2.2; a camera is needed only for live photos (gallery photos also work).
Which software is needed? XAMPP (PHP and MySQL) and Android Studio on Windows; full steps are in the setup guide.
