NAGARA NETRA
AI-Powered Smart City Civic Grievance Platform
Report a civic issue with a photo, voice note, text, and GPS — one ticket, fully structured.
The problem
Civic complaint systems are text-only and frequently route issues to the wrong department. Citizens struggle to describe what they see, and agencies lose the location and visual context they need to act.
What I built
Built a React Native app that ingests four input modalities — photo, voice, text, and GPS — and runs a multimodal pipeline on-device: EfficientNetB0 classifies the issue, BLIP generates a caption, and MuRIL NLP normalizes the multilingual text into a structured grievance.
The result
Every report is auto-tagged and geo-localized, giving agencies a structured, searchable ticket with visual evidence instead of a free-text wall.
Multimodal Pipeline Architecture
Auto-tagged, geo-localized, stored in MongoDB
- Fused image classification + captioning (EfficientNetB0 / BLIP) with MuRIL NLP for multilingual normalization
- Four input modalities — image, text, voice, GPS — merged into a single structured ticket
- On-device ML keeps processing fast and privacy-preserving