Tracking camera T265

Intel® RealSense™ T265 is not a depth camera. It features two fisheye lenses for feature detection, but does not compute dense depth. It is possible to use the image feed from the cameras to compute dense depth, though the results will be poor compared to other Intel® RealSense™ depth cameras as lenses are optimized for wide tracking field of view rather than depth precision, and there is no texture projected onto the environment to aid with depth fill.

All SLAM solutions, and there are many good ones, are limited by the information they receive. It is possible to run host-based SLAM using our D400 series depth cameras – ideally the D435i, however these cameras are optimized for depth accuracy at the expense of field of view – D400 cannot see as much of the world as T265. As such, a SLAM solution based on the information from a single D400 camera will get lost in certain situations when T265 will not. Furthermore, the T265 is optimized for power and latency using its embedded VPU. For these reasons, T265 will succeed in many use cases where D400 based SLAM will fail.
Additionally, when using the T265, no additional resources are required on the platform in terms of compute to perform the SLAM algorithms, which means that tracking with T265 is platform independent, has a low integration cost, and can run on very low compute devices. For some use cases, SLAM on the D435i will be ideal, but for the highest quality tracking, choose the T265. For both quality depth and tracking, use the D415 and the T265 in parallel.

At launch T265 includes support for Windows and Ubuntu. The Host API library is open sourced so customers can port it to any platform they want. It has been run on Android successfully. The FW that runs inside T265 is independent of host platform, so the performance of T265 is also host independent.

Since Intel RealSense T265 computes all tracking data on device, the only hardware requirements are a USB 2 or USB 3 connection that provides 1.5W of power, along with enough memory and compute to boot the T265 and receive the pose data and use it in your application.

The optimization of the algorithm on the VPU prevents it from being shared with other code or uses. Furthermore the thermal envelope of T265 is only large enough for the current workload.

Wheel odometry is the use of sensors to measure how much a wheel turns, it can be used to estimate changes in a wheeled robots position. T265 has wheel odometry support built in, allowing it to use the data from these sensors to refine the position data of the robot. Providing robotic wheel odometer or velocimeter data over USB to TM2 will make the tracking much more robust on wheeled robots, which otherwise can experience many tracking failures. We consider odometer input to be a requirement for robust tracking on wheeled robots.

Intel® RealSense™ T265 can re-localize after kidnapping, providing there are some features in view which are in its internal map. In the case of a completely new environment, it will continue to provide relative pose data until absolute position can be re-established.

Yes, T265 can work both indoors and outdoors. Just like a person, it can be blinded by light that shines straight into its eyes, and it can’t see in absolute darkness.

While well-lit environments are preferable, T265 performs well at light levels as low as 15 lux, and depending on the exact structure of the light can sometimes continue to work at even lower light levels.

An ideal operating environment for the T265 has a reasonable number of fixed, distinct visual features in view. It will perform poorly if the entire field of view contains moving, near field objects such as people. In cases where crowds are expected indoors, it is advised to point the camera upwards, where it can use features on the ceilings to navigate.

No, you can use as many T265 devices in a space as you like.