I trained a YOLOv5 model on Trash Annotations in Context (TACO) dataset using PyTorch.

As a project to do during the winter intersession my first year, I decided to build a robot to pick up trash from beaches. Because I only had a few weeks, I instead built a chassis, wired up a Pi, two roboclaw motor controllers and some planetary DC gear motors to make a chassis that could drive on a beach with dead reckoning. A few semesters ago I picked up the project again during an independent study at Olin. I created a beach anomaly detection script using an adversarial autoencoder and also created a trash detection from video pipeline using YOLO.