This course will survey the state of the art in perception for robotics, which is one of the key enablers to deploying robots in more realistic and unpredictable environments. In this two-day course, we will give an overview of Low-level Vision, Object-level Perception, Tracking and Localization, and mapping techniques, understand the context and situations in which these methods apply, and provide hands-on experience in applying some of these methods to real-world problems.

Course participants will:

• Obtain basic definitions and knowledge of:
  • Low-level Vision: camera models, calibration, feature detection, stereo and range imaging, optical flow
  • Object-level Perception: segmentation, object recognition, pose estimation
  • Tracking and Localization: localization, combinatorial filters, Bayesian inference, Markov and Monte Carlo localization, (multi-target) tracking
  • Perceiving Environments: mapping, occupancy grids, SLAM, SAM and FastSLAM, Structure from Motion, and Visual SLAM.
• Understand the context and situations in which these methods apply.
• Hands-on experience in applying some of these methods to real-world problems.
• Ability to analyze problem and decide which methods apply and what trade-offs are.

• Low-Level Vision
• Tracking/Objects
• Localization
• Robot Mapping

Attendees will each receive copies of handouts and example code.

Student Requirements
Laptop required for in-class exercises

Experience with Java, C++, or C programming is not a requirement, but it will be helpful in gaining the maximum benefit from the hands-on exercises.

• Frank Dellaert
• Irfan Essa
• James Rehg