Debug The Debris

Aaran's roboflow programming with debris datasets
Proof of all ESP-32 parts working except for the camera due to camera visibility.
This is our prototype drone

Our project merges low-cost ESP32 hardware with Roboflow's AI to create computer vision that anyone can access and take advantage from. We built it to provide access to smart technology available for uses like security and object recognition. It is a new, creative way to combat the issue of Marine Debris.

We were simply inspired by the opportunity to bring powerful AI vision at a price point and easy functionality that anyone could use. We used the opportunity of taking one piece of hardware, the very adaptable ESP32 and the easy to use API from Roboflow to make a practical offering to solve real-world problems.

We first created custom firmware for the ESP32 camera for using Wi-Fi to capture images, then analyzed images via the Roboflow API. We conducted rigorous tests and optimization some of which were to balance image quality and limitations of hardware.

Our biggest challenge with the project was first getting over the memory on the ESP32, there were some camera initialization errors as well. Secondly, we figured out an useable manner with API communication while remaining within rate limits. We accomplished this through some very clever coding and optimization. Other hurdles included attempting to pin-point detection with simply making sure plastic wasn't labeled along with the animals within the ocean.

Most importantly we gained an immense amount of experience in memory management for constrained devices and API integration. More importantly we learned how to think through complex hardware and software problems, which we encountered under pressure and became a cohesive team in the process.