Are there benchmarks on the speed (FPS) of the ZED SDK positional tracking module?

Yes. The ZED SDK Positional Tracking module provides pose estimation up to the full camera frame rate, depending on the tracking generation. The ZED SDK includes two positional tracking generations, GEN_1 and GEN_3, designed for different performance and accuracy requirements.

GEN_3 (recommended)

GEN_3 is a feature-based visual-inertial SLAM pipeline that does not rely on depth computation: it is lightweight and supports loop closure and relocalization.

Accuracy (VIO, about 400 m test sequences): mean APE of 0.56 m indoors (warehouse with reflective lights) and 0.29 m outdoors (structured scenes during the day).

GEN_1

GEN_1 is a dense VSLAM approach that relies on depth data, so its maximum frame rate is bound to the compute frame rate of the selected depth mode (see Depth Modes).

GEN_1 is optimized for VIO and is well suited for low-texture or feature-poor environments; GEN_3 is recommended for feature-rich environments, area mapping, and long-term navigation.

These benchmarks were obtained with ZED SDK v5.1.1, ZED X Driver v1.3.2, and a ZED X camera, using the positional tracking sample available on GitHub. Performance also depends on the camera resolution and frame rate, the host platform, and the ZED SDK version.

For the complete and most recent benchmark tables, including accuracy comparisons between all modes, see the Positional Tracking Modes documentation.

Additional resources:

• Positional Tracking overview

• Depth Modes

• How does the ZED 3D perception work?

• Getting started with your ZED camera