Visual Perception for Self-Driving Cars

Overview

Welcome to Visual Perception for Self-Driving Cars, a pivotal third course offered by the University of Toronto within their comprehensive Self-Driving Cars Specialization. This advanced course aims to equip learners with essential skills and knowledge in the realm of autonomous driving perception, focusing on both static and dynamic object detection while exploring a variety of common computer vision techniques used in robotic perception.

Throughout this course, participants will gain proficiency in working with the pinhole camera model, conducting both intrinsic and extrinsic camera calibration, and mastering the art of detecting, describing, and matching image features.

Additionally, learners will have the opportunity to design convolutional neural networks tailored for the automation sector.

Applying these methodologies to critical areas such as visual odometry, object detection and tracking, and semantic segmentation for drivable surface estimation, students will comprehensively understand the foundational elements of self-driving cars' perception systems.

The course's final project challenges participants to create algorithms that accurately identify bounding boxes for scene objects and outline drivable surface boundaries. This hands-on project will allow students to work with both synthetic and real image data, assessing their algorithm's performance through a realistic dataset.

Designed for individuals with a background in computer vision and deep learning, this course requires attendees to have prior programming experience in Python 3.0 and a solid understanding of Linear Algebra.

This course is now available through Coursera and is provided by the National Taiwan University, categorized under Computer Vision, Autonomous Vehicles, and Object Detection courses.

---

Taught by

Steven Waslander

---

Subjects