SVO2 Recording Capabilities

[mukundbala]

Hi, I am using the ZED X/ZED X Mini with the JetsonAGXOrin. I am working on a tool to record video,depth and the corresponding 6D pose data. In the previous version of the SDK (4.0.8), I successfully recorded the video and depth data in the SVO, but found that the poses were not being recorded. As a work around, I saved serialized and saved the poses during data capture. This approach worked quite well for me. It allowed me to retain the compatibility of my collected data with my upstream processing pipeline.

In the latest version, is there a way for me to get VIO poses directly from the SVO2 file? In general, I would need to unpack the svo2 file, extract the data in this format <rgb,depth,pose> before passing it to my upstream processes.

Thanks for your help, let me know if I can better clarify my question!

[Myzhar]

@mukundbala

In the previous version of the SDK (4.0.8), I successfully recorded the video and depth data in the SVO, but found that the poses were not being recorded.

The SVO does not contain depth and pose information. It only saves the RAW stereo camera stream and sensor information.
The new SVO v2 contains sensor data at full rate, and allows you to save custom information as JSON data.

@mukundbala

In the latest version, is there a way for me to get VIO poses directly from the SVO2 file? In general, I would need to unpack the svo2 file, extract the data in this format <rgb,depth,pose> before passing it to my upstream processes.

As I said the VIO information is never saved into the SVO files, the pose of the camera is generated by the ZED SDK at runtime when you set the SVO as the input instead of the live camera.

Please read more here.