Cameras Online
3
Scans Today
0
Open Incidents
0
Critical Priority
0
SLA Breached
0
Verified Closures
0
| Ticket ID | Issue Type | Ward / Location | Priority | Department | Status | Age | SLA Status | Escalation |
|---|
Incidents by Issue Type
Incidents by Ward (Open vs Resolved)
7-Day Hotspot Cameras & Proactive Recommendations
Dynamic AI Watchlist (Zero-Retraining)
Add new civic issue types in plain words (e.g. "burning garbage", "dead animal", "illegal hoarding"). Sent with every AI scan — no model retraining required.
Control Room Access Key
Required for write operations (assigning, resolving, verifying, resetting) to protect the Gemini API quota.
Scanner & AI Diagnostics
Current configuration read from server environment and model metadata:
- Active Vision Model: gemini-2.5-flash
- Minimum Confidence Threshold: 50%
- Persistence Confirmation Threshold: 3 consecutive scans
- SLA Mode: Demo Mode (SLA hours treated as minutes for live evaluation)
Hackathon Demo Utilities
Use these before presenting to initialize historical data or reset for a clean run:
Honesty & Transparency Panel (UKIS 2026)
Real in this Prototype:
- AI detection and classification of garbage, dumping, drain blockage, debris and encroachment on camera frames (Google Gemini vision), reasoning directly from the scene and a short camera-context tag — no manually drawn zones.
- Multi-scan persistence filter against false alarms.
- Explainable priority scoring (severity, size, zone sensitivity, recurrence).
- Department routing, SLA tracking and escalation logic.
- Bilingual work orders (Hindi & English).
- AI before/after closure verification with automatic reopen.
- Incident database, live interactive map and analytics.
Simulated in this Prototype:
- CCTV streams are recorded clips played in the browser (a production system would read RTSP/HLS feeds through a camera gateway).
- The WhatsApp/SMS delivery is a preview.
- Department/ward routing matrix and escalation officers are sample data pending validation with SWSM and ULBs.
- The AI infers footpaths/drains/no-dumping areas from the scene and a one-line camera context tag, rather than from officer-surveyed zone maps.
- SLA timers run in minutes in demo mode.
- Seed historical incidents are marked.
Privacy by design: The AI is strictly instructed never to report or identify people, faces, or number plates. Only issue metadata and evidence frames of civic conditions are stored.