Tutorials Detail Advances in Robotics, AR/VR, and Self-Driving Cars at the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
By Lori Cameron and Michael Martinez

Long Beach, California – The 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) is the premier annual computer vision event comprising the main conference and several co-located workshops and short courses. This year’s conference boasted record-breaking attendance at more than 9,000.

The sheer number of registrants maxed out convention center space forcing organizers to close on-site registration.

Nearly half of all attendees came from industry, while the remaining attendees hailed from academia and elsewhere. The top three fields represented were manufacturing, services, and education. Nearly a third of all attendees came from companies with 10,000 employees or more. A whopping 63 percent of attendees were between the ages of 18 and 34.

Did you enjoy CVPR, co-sponsored by the Computer Society? Then sign up for our Conference Alerts to stay informed of your favorites.

It’s not surprising then that the first two days of the conference saw an abundance of packed rooms for the nearly 120 tutorials scheduled for CVPR. Topics covered a broad range of cutting-edge innovations and developments in computational imaging, visual recognition, image analysis, and deep learning.

The tutorial sessions revealed some of the top fields using computer vision technology today—robotics, virtual and augmented reality, healthcare, and autonomous vehicles. Among industry heavy-hitters was Baidu who boasts the largest autonomous vehicle open source platform in the world.

Check out the highlights below.

Want to read about the workshops going on at CVPR? Click here for more.

“Apollo: Open Autonomous Driving Platform”

tailgating

Dr. Tae Eun Choe, who leads the perception team at Baidu, gave the audience a glimpse into Baidu’s Apollo, the largest open autonomous driving platform in the world, with a full stack of hardware and software developed by the autonomous driving community.

With 130 global partners, Baidu plans to equip manufacturers with the technology to mass produce autonomous vehicles. In fact, nearly 100 vehicles will soon be deployed in the San Francisco Bay area based on the Apollo 3.5 platform.

Baidu attracted applicants to its three competitions at CVPR2019 with prize money totaling 6,300 USD. Competitors were asked to most accurately predict the presence of “traffic-agents”—bicycles, pedestrians, and other vehicles—using three different datasets.

“Perception at Magic Leap”

magic-leap

Magic Leap provided listeners with a deep dive into the four main ingredients in creating an immersive spatial computing platform: head pose tracking, world reconstruction, eye tracking, and hand tracking.

“We blend tech, biology, and creativity to reveal new worlds within our world,” says Magic Leap.

Want to stay ahead of other tech innovations? Sign up for our free digital ComputingEdge magazine.

“Learning Representations via Graph-structured Networks”

graph-structured-networks

Recent years have seen a dramatic rise in the adoption of powerful convolutional neural networks for heavy-duty computer vision tasks. However, these networks don’t adequately model several computer vision properties for more difficult AI tasks: pairwise relation, global context, and processing irregular data beyond spatial grids.

The answer, say researchers, is to reorganize the data to be processed with graphs according to the task at hand while constructing network modules that relate and propagate information across the visual elements within the graphs.

“We call these networks with such propagation modules graph-structured networks. We introduce a series of effective graph-structured networks, including non-local neural networks, spatial propagation networks, sparse high-dimensional CNNs, and scene graph networks,” the researchers say.

network-researchers

Above are the seven featured speakers for the “Learning Representations via Graph-structured Networks” tutorial.

“Towards Relightable Volumetric Performance Capture of Humans”

parameterize-humans1

“Detect, Reconstruct, Track and Parameterize Humans” was just one of the talks held in morning and afternoon sessions that walked the audience through the ins and outs of building, from the ground up, a volumetric capture pipeline for reconstructing, tracking, and texturing of humans in.

“Volumetric (4D) performance capture is fundamental for AR/VR content generation,” says the research team.

This year’s conference was held at the Long Beach Convention Center in Long Beach, California, on 16-20 June 2019.