Multiple Object Tracking using YOLOv8

Published:

The title is self explanatory. I made this project as a part of our data collection pipeline for CrowdSurfer. CrowdSurfer required a lot of data and we decided to collect all the data using our custom wheelchair inside the IIIT Hyderabad campus. MOT using YOLOv8 works with an open-set, meaning it detects any obstacle with a relatively high degree of confidence. I made a ROS implementation so that it can easily work with the other components of the stack.

Features

  • Open-Set
  • Interfaced with RealSense D455 ROS
  • High degree of confidence

Technologies Used

  • Computer Vision
  • PyTorch
  • ROS Noetic

Link to Project Source Code
View the project on YouTube