Genz Hackathon

The prototype drone, which contains a breadboard, some jumper wires and also propellers. This is merely a prototype.

Inspiration We were inspired by the potential of computer vision and IoT technology to create practical solutions for real-world problems. The combination of ESP32's affordability and Roboflow's powerful AI capabilities presented an opportunity to build an accessible smart camera system that can be deployed in various scenarios for marine debris.

What it does Our system captures images using an ESP32 camera module, processes them through Roboflow's computer vision API for object detection and recognition, and then makes decisions or alerts based on the analyzed content. The system can identify specific objects, count items, or detect presence in a monitored area.

How we built it Hardware: ESP32 WROVER module with OV2640 camera Firmware: Custom Arduino code handling camera initialization, Wi-Fi connectivity, and image capture Computer Vision: Integration with Roboflow's API for image analysis Connectivity: HTTP requests to send images and receive analysis results Response System: Logic to act on the analysis results (e.g., send alerts, count objects)

Challenges we ran into Camera initialization issues due to memory allocation problems Inconsistent Wi-Fi connectivity affecting API communication Optimizing image quality vs. file size for efficient API transfers Managing API rate limits and response handling Synchronizing hardware operations with network requests

Accomplishments that we're proud of Successfully resolving the camera initialization error through memory optimization Creating a reliable connection between ESP32 and Roboflow's API Achieving near real-time object detection on limited hardware Being close to finishing with a nice drone prototype to present to the judges

What we learned Advanced memory management techniques for constrained devices Optimizing neural network inference for edge devices Team collaboration on both hardware and software components About each other, and how we can use each other efficiently to finish such minor errors and issues

What's next for Genz Hackathon Implement additional computer vision models for different applications Add local processing capabilities to reduce dependency on cloud APIs Develop a mobile app for remote monitoring and control