Why is my application frozen when using the NEURAL depth mode?

The NEURAL depth mode uses an Artificial Intelligence model to extract depth information:

sl::InitParameters::depth_mode = sl::DEPTH_MODE::NEURAL

This AI model, like the other AI model available with the Object Detection module of the ZED SDK, the first time that is used must be downloaded from the Stereolabs server and optimized for the GPU model installed on the host device. The ZED SDK automatically performs these operations.

The download is required because each AI model takes a big amount of disk space; including them all in the ZED SDK installer would result in a large amount of data to be downloaded with a single download.

The optimization step instead allows getting the full performances from the AI engine of the ZED SDK by adapting the model to the architecture of the Nvidia GPU card installed on the host machine.

This procedure must be executed only one time after installing a new version of the ZED SDK.

Downloading and optimizing the AI model takes a few minutes; if this phase is not correctly handled by your application, it will result in a not responding user interface.

We suggest downloading and optimizing the NEURAL depth AI model before starting your application by running the ZED Diagnostic tool in the command line interface, with the option `--ais X`:

  • Windows
    ZED Diagnostic.exe --ais 8
  • Linux
    ZED_Diagnostic --ais 8

Note: the NEURAL depth mode is available starting from ZED SDK v3.7.0.

Note: starting with the ZED SDK v3.7.1 you can download and optimize the NEURAL depth AI model by using the ZED Diagnostic option `--nrlo`, or by using the GUI interfaces of the `ZED Depth Viewer` and `ZED Diagnostic` tools.