Inspiration
The genesis of EmpowAir emerged from a deeply personal understanding of a universal concern - the fear that accompanies many women on their daily journeys. As a DS immersed in AI development, I witnessed a stark contrast, while we possess sophisticated technology, women still clutch their keys between their fingers while walking home, share their live locations with friends, and make anxiety-filled phone calls during late-night commutes.
Current surveillance systems, despite their prevalence, remain passive observers. They lack the intelligence to:
- Differentiate between casual interaction and potential threats
- Understand the contextual nuances of a situation
- Provide real-time intervention capabilities
- Analyze patterns and predict potential risk areas
This gap became evident through conversations with:
- Women working late shifts in urban areas
- Healthcare professionals commuting during odd hours
- Students navigating campus areas after evening classes
- Professional women balancing career demands with safety concerns
The moment of realization came when combining three observations:
- The anxiety in a friend's voice during our daily "I'm walking home alone please accompanying me through this video call"
- The potential of AI to understand and analyze human behavior
- The untapped capability of drone technology for proactive surveillance
EmpowAir emerged as a solution that combines:
- YOLO's ability to detect physical threats in real-time
- Gemmini-1.5-flash
- Geospatial intelligence to analyze environmental risks
The goal wasn't just to create another surveillance system, but to develop an intelligent guardian that:
- Watches when she walks alone
- Understands the difference between normal activity and potential threats
- Provides real-time analysis and response
- Creates safer pathways through technology
Because in 2024, no woman should have to:
- Plan their career choices around daylight hours
- Feel afraid walking home after pursuing their dreams or working their jobs
- Carry the constant burden of vigilance
- Consider safety before opportunity
EmpowAir represents more than a technical solution; it's a step toward a world where:
- Career choices aren't limited by commute times
- Educational opportunities aren't restricted by campus locations
- Professional growth isn't hindered by safety concerns
- Women can focus on their achievements rather than their safety
Through the synergy of AI technologies, we're not just detecting threats - we're working to eliminate the need for those anxiety-filled phone calls, the rushed walks home, and the constant looking over one's shoulder. Because every woman deserves to walk with confidence, chase her dreams without fear, and know that technology has her back, quite literally, from above.
What it does
EmpowAir (Simulation Version 1.0) is a proof-of-concept drone surveillance system that demonstrates the potential of integrating three key AI technologies for women's safety. Currently implemented as a simulation, this project showcases:
π€ YOLO Violence Detection
- Simulated real-time monitoring and threat detection
- Recognition of multiple violence categories in test scenarios
- Weapon detection capabilities in controlled environments
π§ Gemma LLM Analysis
- Experimental contextual situation understanding
- Risk assessment based on simulated scenarios
- Generation of actionable safety recommendations
πΊοΈ Geospatial Intelligence
- Mock location-based context analysis
- Simulated terrain and environment assessment
- Theoretical access point and escape route mapping
How we built it π§
As a data scientist venturing into full-stack development, this simulation was built systematically:
Core AI Layer
- YOLO model configured for violence detection scenarios
- Gemma LLM integration for analytical processing
- Custom geospatial processing module using test data
Backend Architecture
- FastAPI server handling simulated drone feeds
- PostgreSQL with PostGIS for spatial data management
- Pipeline processing simulated real-time data
Frontend Development (The Learning Curve!)
- React-based dashboard for data visualization
- Interactive simulation controls
- Real-time update system for test scenarios
Challenges we ran into
Solo Full-Stack Journey
- Transitioning from data science to frontend development
- Self-teaching React ecosystem during development
- Balancing backend expertise with frontend learning
AI Integration Hurdles
- Synchronizing multiple AI models in simulation
- Optimizing processing for mock video feeds
- Creating reliable detection scenarios
Technical Constraints
- Resource management for simultaneous AI operations
- Simulating real-time data streams effectively
- Implementing robust state management
Accomplishments that we're proud of π
AI System Integration
- Successfully coordinated YOLO and Gemma in simulation
- Achieved efficient processing in test scenarios
- Demonstrated decent threat detection capabilities
Frontend Development Victory
- Created functional dashboard from scratch (lol proud of me, thanks Gemma for guiding me ofc)
- Implemented visualization of simulated data
- Designed intuitive interface for demo purposes
System Capabilities
- Demonstrated low-latency processing potential
- Reliable detection in test scenarios
- Effective analytical pipeline
What we learned
Full-Stack Development
- Foundations of frontend architecture
- React component development
- Modern state management approaches
- Basic UI/UX principles
System Design
- Real-time system architecture
- API design patterns
- Data flow optimization
- Performance considerations
AI Implementation
- Multi-model integration strategies
- Simulation-based AI testing
- Performance optimization methods
What's next for EmpowAir π
From Simulation to Reality
- Integration with actual drone hardware
- Real-world testing and validation
- Performance optimization in live conditions
AI Enhancement
- Training specialized Gemma model variants
- Implementing advanced pattern recognition
- Improving contextual understanding
Edge Computing Integration
- Moving from simulation to edge deployment
- Optimizing for real-world latency
- Enhancing real-time capabilities
Dashboard Evolution
- Implementing live geospatial mapping
- Developing advanced analytics tools
- Creating comprehensive monitoring systems
Note: The current version is a simulation demonstrating the potential of this technology. Our next steps focus on transitioning from simulation to real-world implementation while maintaining our commitment to enhancing women's safety through technology.
*This project is developed to demonstrate the potential of AI-driven surveillance systems in enhancing women's safety.
Log in or sign up for Devpost to join the conversation.