Skip to main content

Intelligent AI Detective

Table of Contents

How?
#

  • Designed, generated data, and trained a CNN to detect alphanumeric characters from an image.
  • Utilized OpenCV to process and filter images from a simulated camera feed in Gym Gazebo.
  • Developed imitation learning model and classical control algorithms to maneuver a robot on a simulated road switching between states avoiding obstacles, following leaders, and obeying simplified traffic laws.
  • Implemented a homography-based tracker using SIFT to highlight objects in a video feed.

Results
#

  • Agent was able to successfully maneuver across the entire world and predict all clues correctly.
  • Learned how to use ROS and Linux (Xubuntu).

Related

Autonomous Racing Robot
UBC AeroDesign