This project is an offline object detection system that uses a DIY drone and a Raspberry Pi to identify objects without an internet connection. It is designed to work in remote or restricted areas where cloud services are unavailable.
Team
Sponsors & Advisors
How It Works
Aerial Capture: The process begins with a quadrotor drone that captures live footage from its surroundings while in flight.
Data Transmission: This footage is broadcast wirelessly through the drone's built-in Wi-Fi module.
Local Processing: A Raspberry Pi 5 hub receives the video stream and uses a locally installed YOLO deep learning model to analyze every frame.
Real-Time Identification: The AI instantly identifies objects—such as people or vehicles—and overlays bounding boxes and confidence scores onto the video.
Offline Output: The final processed feed is displayed on a touchscreen, providing high-tech surveillance without ever needing a cloud or internet connection.
Future Improvements
- Processing Power: Adding a dedicated GPU or TPU for faster, more accurate AI models.
- Automated Reports: Creating logs that summarize what was detected and when to save review time.
- Direct Streaming: Using open-source drones to stream video directly to the Pi without a middle device.