# Save floods

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Dhruva Wani](https://cerebralvalley.ai/u/dhruvawani17)
- **GitHub:** https://github.com/wanidhruva/aquatrace
- **Demo video:** https://youtu.be/e6eK6dnsXpI
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2625

AquaTrace — See. Decide. Act.

Multimodal Flood Intelligence & Early Warning System

AquaTrace is an AI-powered flood intelligence and early-warning system designed to detect developing flood threats and deliver actionable alerts before they turn into disasters.

Flood-prone regions often lack continuous, localized monitoring. Traditional systems depend on dedicated monitoring stations, manual observation, or periodic weather updates, which may fail to capture rapidly changing conditions.

AquaTrace addresses this challenge by transforming existing CCTV infrastructure into intelligent flood-monitoring systems. It combines computer vision, CCTV footage, rainfall and weather information, environmental data, and sensor readings to provide a more comprehensive understanding of potential flood risks.

How It Works

1. SEE — Intelligent Detection

Uses computer vision to analyze CCTV footage and identify:

Water-covered areas and changes in water-area ratio.

Waterline displacement and rising water levels.

Flood expansion and water movement patterns.

Unusual visual changes indicating potential flooding.

It tracks changes across consecutive observations to understand how conditions evolve over time.

2. DECIDE — Multimodal Intelligence

Combines visual observations with rainfall, weather conditions, sensor readings, and temporal trends to assess potential threats.

The system follows three progressive risk stages:

Watch: Initial abnormal conditions detected.

Escalating: Continued water-level increases and supporting risk indicators.

Critical: Potentially severe conditions requiring immediate attention.

Nearby camera feeds can also provide additional verification and context.

3. ACT — Evidence-Based Alerts

A centralized dashboard presents monitoring feeds, water-level changes, environmental indicators, risk progression, alert status, confidence information, and supporting visual evidence.

Key Innovations

Leverages existing CCTV infrastructure instead of depending entirely on dedicated monitoring hardware.

Combines computer vision with environmental and sensor-based information.

Uses temporal analysis to track how flood conditions develop.

Applies progressive risk assessment instead of simple water detection.
Supports scalable monitoring across multiple locations.

Technology

AquaTrace integrates computer vision, image analysis, temporal tracking, multimodal data processing, risk assessment, and dashboard visualization. Its architecture is designed to support advanced multimodal AI capabilities from Meta's ecosystem.

Vision & Impact

The vision is to transform ordinary surveillance infrastructure into an intelligent, distributed flood-monitoring network. AquaTrace aims to improve early-warning capabilities in vulnerable and under-monitored regions, helping communities and authorities recognize potential threats earlier and make better-informed decisions.

AquaTrace — Detecting threats before they become disasters.

## More from Meta Global AI Developer Hackathon

- [Danna Team](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2622)
- [PolicyProof](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2623)
- [Promptimize](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2624)
- [Maya's era](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2626)
- [Along](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2627)
- [Nextwaves Industries](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2628)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2625. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt