Built for Hackathon 2026 — Turning Data into Defenses. Crafted with passion by Team CodeCrust
CrimeMap transforms fragmented regional crime data into a live, actionable safety intelligence grid. Designed for both law enforcement and everyday citizens, it fuses open-source intelligence, interactive heatmap visualization, a 4-state verified citizen reporting pipeline, and a passive under-reporting detection layer into a single decision-ready platform.
| Feature | Description | |
|---|---|---|
| 📡 | Ranked OSINT Intelligence Feed | Real-time crime news scored by source trust + incident severity and presented as top-signal articles |
| 🌐 | Interactive Heatmap | National crime density visualization with district-level drill-down, safety scores, and regional ranking |
| 🧠 | Passive Under-Reporting Detection | Weighted complaint signal algorithm flags jurisdictions where citizen intel far exceeds official records |
| 🏛️ | Admin Responsiveness Intelligence | Tracks and visualizes governance bottlenecks — districts with low complaint clearance rates get |
| 🙋 | 4-State Citizen Reporting Lifecycle | UNVERIFIED → UNDER_REVIEW → VERIFIED → REJECTED with full audit trail, evidence upload, and privacy controls |
| ⚙️ | Full-Stack Architecture | FastAPI backend + React frontend, JWT-secured, file-serving for evidence, production-ready |
The platform ingests live news signals, scores them, and surfaces only the highest-signal articles.
- Real-time severity scoring — articles are ranked using keyword intensity (
murder,rape,robbery) and source trust tiers - Top-6 intelligence filter — only the six highest-ranked articles are presented, eliminating noise
- Intelligent caching — 5-minute TTL cache protects API rate limits and ensures sub-second responsiveness
Note: News data is powered by NewsAPI (free tier). High-frequency usage may encounter rate limits.
- Dynamic hotspot detection — crime density rendered across states and districts in real time with animated pulse markers
- Safety Index & Regional Rank — each district popup shows a calculated safety score (0–100) and its statewide rank
- Dual badge overlay — map markers carry a red
!for under-reporting anomalies and an orange⚠️for low admin responsiveness, always visible without clicking - Decision support overlays — highlights "Actionable Intelligence Zones" and "Community Gap Alerts"
- Weighted complaint signal — complaints are scored by status:
UNVERIFIED=1,UNDER_REVIEW=2,VERIFIED=3 - Anomaly detection — districts where signal strength significantly exceeds official crime scores are automatically flagged as potential under-reporting zones
- Zero ML required — entirely deterministic algorithm, no external model dependencies
- Response rate computation — system calculates the ratio of validated-to-pending intel per jurisdiction
- Visual governance scoring — districts with a critically low response rate are marked with a persistent
⚠️badge directly on the map marker - InfoWindow details — clicking a flagged district reveals the exact response rate percentage and an explanation of the bottleneck
- 4-state verification lifecycle —
UNVERIFIED → UNDER_REVIEW → VERIFIED → REJECTEDwith color-coded map pins (Purple/Blue/Green/Red) - Soft-enforced admin flow — admins can fast-path to
VERIFIEDdirectly or step through review; all transitions are timestamped - Pending-time visibility — admin table shows "Pending for X hours/days" for unresolved reports, creating accountability pressure
- Evidence upload — citizens can attach image evidence; files are stored server-side and rendered in the admin Intelligence Briefing
- Privacy-first reporting — users toggle whether their detailed description is shared on the public map; admins always retain full access
- My Reports dashboard — citizens track their submission history with live lifecycle status, age tracking, and reasoning for rejections
| Layer | Technology |
|---|---|
| Frontend | React 18, Vite, Tailwind CSS, Lucide Icons, Axios |
| Backend | FastAPI (Python), Uvicorn, Python-Multipart |
| Auth & Security | JWT Authentication, PBKDF2 Password Hashing (Passlib) |
| Data Engine | Pandas — regional aggregation and analytical compute |
| Storage | JSON flat-file DB (no external database dependency) |
| File Serving | FastAPI StaticFiles — persisted evidence uploads via /uploads |
- Node.js v18 or higher
- Python 3.10 or higher
# 1. Navigate to the backend directory
cd backend
# 2. Create and activate a virtual environment
python -m venv venv
source venv/bin/activate # Windows: venv\Scripts\activate
# 3. Install dependencies
pip install -r requirements.txt
# 4. Configure environment variables
cp .env.example .env
# Edit .env and fill in your values:
# NEWS_API_KEY=your_newsapi_key_here
# JWT_SECRET=your_jwt_secret_here
# 5. Start the development server
python -m uvicorn app.main:app --reloadBackend will be live at http://localhost:8000
# 1. Navigate to the frontend directory
cd frontend
# 2. Install dependencies
npm install
# 3. Start the development server
npm run devFrontend will be live at http://localhost:5173
- Register or log in as a standard user
- Explore the Intelligence Feed — browse ranked OSINT crime articles filtered by source trust and severity
- Check the Safety Map — view verified hotspots, under-reporting alerts (
!), and governance warnings (⚠️) on the live heatmap - Submit a Report via the Reporting Desk — attach evidence images, optionally keep your description private, and track the full verification lifecycle under My Reports
- Log in using specialized admin credentials
- Access the Intelligence Desk — review incoming citizen reports with full reporter details, GPS coordinates, and attached evidence
- Manage the Lifecycle — step reports through
Under Reviewor fast-pathVerify; reject with optional reasoning - Monitor Accountability — the "Pending for X" column and responsiveness analytics immediately surface governance backlogs
- No heavy databases — all data persisted in JSON files (
complaints_db.json,auth_db.json) for zero-dependency deployment - In-memory intelligence — all enrichment metrics (admin responsiveness, signal strength) computed on-demand from
_processed_datacache - Stateless backend — cache is invalidated and re-enriched on each complaint state change, keeping analytics always fresh
- Privacy by design — citizen PII (name, phone, email) is stored securely but masked from all public-facing API surfaces and map overlays
Team CodeCrust — Hackathon 2026
| Name | GitHub |
|---|---|
| Shreya Nipunge | @Shreya-nipunge |
| Kirtan Devadiga | @Kirtan-pc |
| Mayuri Gade | @mayurigade-hub |
| Shreya Dandekar | @shreyadandekar |
| Lata Choudhary | @latachoudhary18 |
This project is licensed under the MIT License — see the LICENSE file for details.