CleanSight
CleanSight is a full-stack, AI-powered smart waste reporting and municipal management platform connecting citizens, volunteers, municipal staff, and city administrators across four role-gated dashboards. The platform features a two-stage computer vision ML pipeline in Python (FastAPI, TensorFlow MobileNetV2, and PyTorch) for binary trash validation and 4-class waste material classification (glass, paper, plastic, mixed) with confidence-based triage and human-in-the-loop review. Built with a React 18 + TypeScript frontend (Vite, Tailwind CSS) and a Node.js + Express backend with MongoDB Atlas, it integrates MapLibre GL for geospatial vector mapping, Firebase Authentication with 4-tier RBAC, BullMQ/Redis for asynchronous ML job queueing, and Server-Sent Events (SSE) for real-time administrative activity streaming.
14
Total Technology
6
Key Features
Technologies Used
Key Features
- Two-Stage AI Computer Vision Pipeline — Automated binary trash validation (MobileNetV2) and 4-class waste material classification (PyTorch) with confidence-based triage and human-in-the-loop review
- Role-Gated Operational Dashboards — 4 dedicated dashboard interfaces tailored for Citizens (reporting & badges), Volunteers (task discovery & proof submission), Municipal Staff (triage & dispatch), and Admins (executive KPIs & monitoring)
- Interactive Vector Geospatial Mapping — MapLibre GL JS integration with spatial clustering, bounding-box viewport queries, GPS auto-detection, and distance calculation for pending cleanups
- Real-Time Admin Activity Streaming — Server-Sent Events (SSE) stream delivering live city cleanup events and operational updates directly to admin control centers
- Asynchronous Job Queue & Scalable Backend — BullMQ with Redis to decouple compute-heavy AI image analysis from Express REST API response cycles
- Enterprise Security & Audit Logging — Firebase Authentication, 4-tier server-side RBAC, Helmet security headers, rate limiting, and comprehensive administrative audit trails