Inspiration
The inspiration for our project stemmed from the critical need to enhance navigation and location tracking capabilities, especially in environments where GPS signals may be unreliable or unavailable, such as dense urban areas or regions with signal interference. Drones, being increasingly utilized for various applications including surveillance, delivery, and mapping, often face challenges in GPS-denied areas. We sought to develop a solution that leverages machine learning to accurately identify coordinates from aerial photos in real-time, thus providing a robust alternative to traditional GPS-based systems.
What it does
Our project, DroneNav, employs machine-learning techniques to track drone locations in GPS-denied environments using aerial shots. It accurately identifies coordinates and provides real-time location information, offering a reliable alternative to GPS-based systems.
How we built it
- Preprocessed the source images to extract relevant frames and generate corresponding ground truth coordinates.
- Utilized state-of-the-art object detection models for identifying locations in aerial shots.
- Implemented dissection algorithms to maintain continuity and accuracy in location estimation over the space.
- Integrated Deep Learning models into the backend pipeline for seamless inference.
- Designed an interactive demo webpage with a topographical image interface for users to hover over regions of interest
- Integrated the front end with the backend API to fetch and display accurate coordinates and location information in real-time.
- Hosting the implementation online for demo and live preview of the functionality.
Challenges we ran into
- Missing Location Data: We manually generated the location data based on matching it to Google Maps and we dissected the base image into patches for individual detection.
- Real-time Processing: Optimizing the backend methods/infra for efficient real-time image processing and inference required careful resource allocation and performance tuning.
- User Experience: Balancing functionality with simplicity in the frontend interface to provide a seamless user experience while conveying complex location data with the generated location info.
Accomplishments that we're proud of
High accuracy in identifying and tracking drone locations, especially in urban locations.
What we learned
Data preprocessing techniques, ML model pipelines, and online hosting mechanisms.
What's next for DroneNav: GPS-Free Drone Location Tracking
- Enhanced Accuracy: Continuously refine and optimize machine learning models to improve location accuracy and robustness in diverse environments.
- Expanded Feature Set: Integrate additional features such as altitude tracking, trajectory prediction, and obstacle detection to enhance overall drone navigation capabilities.
- Deployment and Integration: Work towards deploying DroneNav as a practical solution for drone navigation in real-world scenarios, collaborating with industry partners and regulatory bodies for seamless integration and adoption.
Log in or sign up for Devpost to join the conversation.