Perception and Learning for Autonomous Driving - Part 1

Explore a comprehensive tutorial on perception and learning for autonomous driving in this first part of a two-part series. Delve into robust multi-modal perception, sensor suites, and the application of deep neural networks in autonomous vehicles. Examine the evolution of convolutional neural networks, their architectures, and their role in classification, detection, and segmentation tasks. Investigate challenges in object detection, including sliding window detection and the PASCAL VOC Challenge. Learn about R-CNN and Fast R-CNN training techniques, as well as multi-task loss functions. Gain valuable insights into the mathematical challenges and opportunities in autonomous vehicle technology from Jana Kosecka of George Mason University.

Intro
Robust Multi-modal Perception
Sensor Suite
Perception and Learning for Autonomous Driving
Deep Neural Networks / DNNS
Convolutional networks
Convolutional Neural Networks
Single Layer Architecture
Deep Learning
Classification, Detection, and Segmentation
Architectures Evolution
Factorized convolution
Parameters and computation
Residual Networks
Classification Detection, and Segmentation
Challenges of object detection?
Conceptual approach: Sliding window detection
PASCAL VOC Challenge (2006-2012)
R-CNN details
Fast R-CNN training
Multi-task loss