Supported Hardware
RGB-D Imagers currently supported:
- Kinect v1 (XBOX, not Kinect for Windows)
- Mesa SwissRanger 4500
- Stereo Camera (Point Grey BlackFly - USB or IP)
- Kinect v2
The main requirement for each compute node: Linux-compatible x86 hardware capable of running Ubuntu 14.04 and the Robot Operating System (ROS). Most likely, one compute node will be needed per imager. Our deployment experience suggests that modern Ubuntu-compatible machines with i7 CPUs and at least 8 GB RAM constitute a good setup for each node. Running on pre-2011 hardware is not recommended. OpenPTrack will not compile on a machine with less than 2.7GB RAM.
Faster CPU(s) can mean better tracking performance, and may also enable running more than one detection process (i.e., more than one sensor) per node. Do not run any node close to 100% CPU utilization, or tracking performance suffer will most likely suffer.
Kinect v2 support requires a CUDA-capable GPU. The NVIDIA GPUs that have been tested so far: NVidia GeForce 650, 660, 670, 740, 750, 760, 770, 840, 850, 860, and 870. Recommended: 384 CUDA cores (or more).
The Deployment Guide contains additional hardware recommendations, in the Tested Hardware section.
- Detection and tracking hosts: Ubuntu 14.04.
- Consumers: Any platform that can receive UDP datagrams.
If you are working with OpenPTrack v2, use this wiki instead.
- Time Synchronization
- Pre-Calibration Configuration
- Intrinsic Calibration
- Camera Network Calibration
- Person Tracking
- Object Tracking
- World Coordinate Settings
- Tips and Tricks
- Tested Hardware
- Network Configuration
- Imager Mounting and Placement
- Calibration in Practice
- Quick Start Example
- Imager Settings
- Manual Ground Plane
- Calibration Refinement
How to receive tracking data in:
Notes on contributing to OPT.