YoVDO

Label Efficient Visual Abstractions for Autonomous Driving

Offered By: Andreas Geiger via YouTube

Tags

Autonomous Vehicles Courses Machine Learning Courses Computer Vision Courses Autonomous Driving Courses

Course Description

Overview

Explore a keynote presentation on label-efficient visual abstractions for autonomous driving. Delve into the trade-offs between annotation costs and driving performance in semantic segmentation-based approaches. Learn about practical insights for exploiting segmentation-based visual abstractions more efficiently, resulting in reduced variance of learned policies. Examine the impact of different segmentation-based modalities on behavior cloning agents in the CARLA simulator. Discover how to optimize intermediate representations for driving tasks, moving beyond traditional image-space loss functions to maximize safety and distance traveled per intervention.

Syllabus

Introduction
Two dominating paradigms to selfdriving
Direct perception
Conditional forints learning
Intermediate representations
More Related Findings
What is a good visual abstraction
InputOutput
No Crash Benchmark
Identifying Relevant Classes
Results
Qualitative Results
Summary
Dataset Overview
Illustrations


Taught by

Andreas Geiger

Related Courses

4.0 Shades of Digitalisation for the Chemical and Process Industries
University of Padova via FutureLearn
A Day in the Life of a Data Engineer
Amazon Web Services via AWS Skill Builder
FinTech for Finance and Business Leaders
ACCA via edX
Accounting Data Analytics
University of Illinois at Urbana-Champaign via Coursera
Accounting Data Analytics
Coursera