Overview
The CycleSafely project is actively developing new features focused on visualization, mobile optimization, and real-time safety monitoring. Current work integrates advanced sensors with user-friendly interfaces to provide cyclists with immediate feedback about their safety on the road.
Primary Focus Areas
Our current development efforts concentrate on creating a complete end-to-end system that cyclists can use in real-world scenarios, from data collection through real-time monitoring to post-incident analysis.
Active Development Tasks
Visualization Platform Active Development
Building a comprehensive visualization system that combines multiple data sources:
- Tracked Vehicle Data: Display vehicle trajectories, speeds, and distances in an intuitive interface
- Geolocated Collision Statistics: Overlay historical collision data on map views to identify dangerous locations
- Road Width Analysis: Integrate road geometry data to assess available space for safe passing
- Interactive Analysis Tools: Allow users to explore their rides and understand safety patterns
Smartphone Camera Input Processing Active Development
Developing lightweight vehicle detection directly from smartphone cameras:
- YOLO3D Integration: Vehicle detection from photos using optimized YOLO3D models
- Monocular Depth Estimation: Estimating distances without dedicated depth sensors
- Real-time Processing: Achieving acceptable frame rates on mid-range smartphones
- Battery Optimization: Balancing detection accuracy with power consumption
Multi-Sensor Device Integration Active Development
Creating a comprehensive sensor package mounted on bicycle handlebars:
- Smartphone Mount: Secure mounting solution with optimal camera positioning
- GPS + IMU Registration: High-accuracy positioning and motion tracking
- ORB-SLAM3 Integration: Visual SLAM for robust localization and mapping
- Sensor Synchronization: Precise timestamping across all input devices
Advanced LIDAR System Active Development
Integrating high-precision sensors for research-grade data collection:
- Livox Mid-360 LIDAR: 360° point cloud capture for accurate distance measurements
- Insta 360 Camera: Panoramic photo capture for comprehensive scene documentation
- Relative Speed Tracking: Computing vehicle speeds relative to cyclist
- Absolute Speed Computation: Calculating actual vehicle speeds using GPS+IMU data
Real-Time Safety Display Active Development
Developing an intuitive display system for immediate safety feedback:
- Tablet Display: Mounted on bicycle rack for easy visibility
- Distance Monitoring: Real-time indication of whether 1.5m safety distance is maintained
- Speed-Based Alerts: Different thresholds for vehicles traveling at different speeds (30km/h threshold)
- Visual + Audio Warnings: Multi-modal alerts to ensure cyclist awareness
Mobile Platform Optimizations Active Development
Porting BA Kozonits optimizations to mobile deployment:
- Model Quantization: Adapting optimized models for iOS and Android
- User Interface Design: Creating intuitive, glanceable displays for cyclists
- Background Processing: Ensuring continuous operation while cycling
- Storage Management: Efficient buffer management for video recording
Anonymized Data Upload System Planning
Building infrastructure for crowdsourced safety data collection:
- Automatic Anonymization: Removing personally identifiable information before upload
- Incident Detection: Automatically identifying and storing close-pass events
- Video Buffer Storage: Saving last N seconds of video when incidents occur
- Post-Mortem Analysis: Allowing users to review and understand dangerous situations
Technical Challenges
Real-Time Performance
Achieving low-latency processing on mobile devices while maintaining accuracy:
- Balancing detection quality vs. speed
- Optimizing for various smartphone hardware
- Minimizing power consumption
Robust Tracking
Maintaining stable vehicle tracking in challenging conditions:
- Handling occlusions and partial views
- Dealing with varying lighting conditions
- Tracking at different speeds and distances
User Experience
Creating interfaces that are useful without being distracting:
- Designing glanceable safety indicators
- Avoiding alert fatigue
- Ensuring visibility in sunlight
Data Quality
Ensuring collected data is useful for analysis:
- Accurate sensor calibration
- Precise GPS+IMU fusion
- Reliable distance measurements
Next Steps
Upcoming milestones for the CycleSafely project:
- Field Testing: Real-world evaluation with cyclists in urban environments
- User Studies: Gathering feedback on display design and alert effectiveness
- Dataset Release: Publishing anonymized cycling safety data for research
- Mobile App Release: Beta testing of smartphone-only version
- Integration Testing: Full system evaluation with all sensors
Get Involved
The CycleSafely project welcomes contributions in various forms:
- Student Projects: Bachelor's and Master's thesis opportunities available
- Data Collection: Help expand the dataset with diverse cycling scenarios
- Testing: Participate in field trials and provide feedback
- Development: Contribute to computer vision, mobile optimization, or UI design
Bonus: Exceptional implementations can earn a €1000 bonus!