Features¶
Implemented Features¶
Mobile Application¶
- Waste Identification: On-device waste classification using TensorFlow Lite models with 84% mAP50 and 76% mAP50-95 accuracy
- User Authentication: Email/password, Google Sign-In, and phone verification
- Community Marketplace: Create and browse posts for giveaways and requests
- Private Messaging: Chat functionality for post-related communications
- Event Management: Create events and RSVP as participants
- Location Services: Find and manage recycling centers using Google Maps
- Gamification: Earn points and unlock badges for recycling activities
- Push Notifications: Real-time notifications via Firebase Cloud Messaging
Backend API¶
- GraphQL API: Complete API for all mobile features using NestJS
- Database: MongoDB with Prisma ORM for data management
- Authentication: JWT-based auth with Firebase integration
- Geospatial Queries: Location-based searches using MongoDB geospatial indexes
Machine Learning¶
- Model Training: Python utilities for training waste classification models using YOLOv8
- Dataset Management: Scripts for preparing and merging training datasets
- Model Performance: YOLO waste detection model with 84% mAP50, 76% mAP50-95, 81% precision, and 81% recall
- Waste Categories: Detects 8 categories (cardboard, glass, metal, paper, plastic, trash, biological, battery)
- Model Size: ~2.7 MB TFLite model optimized for mobile deployment
Last Updated: December 2025