Efficient Human-in-the-Loop Computer Vision Algorithms to Create Datasets of Rare Traffic Events from Video
Walter Lasecki, PhD
Assistant Professor, Electrical Engineering and Computer Science, U-M College of Engineering
January – December 2017
Jason Corso, PhD
Associate Professor, Electrical Engineering and Computer Science, U-M College of Engineering
To create hybrid-intelligence pipelines for efficiently reconstructing 3D scenes from 2D video using crowdsourcing and computer vision.
We will create tools and models that allow us to effectively (and efficiently) trade off human and machine intelligence for efficient fine-grained annotation of traffic scenes. This will enable large-scale use of traditional 2D video for simulation-based training of autonomous vehicles.
A proof-of-concept pipeline for reducing overall cost/human effort needed to extract fine-grained parameters from a scene in 2D video.