Communications of the ACM
Tracking Humans in 3D With Off-the-Shelf Webcams
Computer scientists have developed a system that requires only a web camera to capture a person's movements digitally in three dimensions.
Credit: Oliver Dietze
Researchers at the Max Planck Institute for Computer Science in Germany have developed VNect, a system for capturing human movements digitally in three dimensions (3D) in real time using a single video camera.
The system also can estimate the 3D pose of a person acting in a pre-recorded video, offering new applications in character control, virtual reality, and ubiquitous motion capture with smartphones.
VNect is based on a convolutional neural network that can calculate the 3D pose of a person from the two-dimensional information of the video streams. The new system avoids wasting computations on image regions that do not contain a person.
The neural network was trained using tens of thousands of annotated images during the machine learning process.
Although the accuracy of the pose estimation is slightly lower than the accuracy obtained with multi-camera or marker-based pose estimation, the researchers believe the technology will further mature and be able to handle increasingly more complex scenes.
From Saarland University
View Full Article
Abstracts Copyright © 2017 Information Inc., Bethesda, Maryland, USA
No entries found