Hi. I’m new to using ZED and their svo format.
I’m using ZEDfu for offline 3D reconstruction, which works but leaves many artifacts. I’ve extracted the depth frames to run preprocessing algorithms on them and wonder if I can rewrap it into the svo format to use ZEDfu (or if ZEDfu could use my depth images instead of the ones in the original svo file).
If that isn’t possible, what would you recommend doing?
Thanks in advance
Hi,
The SVO file does not contain the depth images, but only the RGB images, the sensors data (IMU, etc) and some metadata. The depth images are computed at runtime during the replay of the SVO file.
Didn’t know that. Is there ZED software that could take in custom depth images?
No, it is not possible to ingest either RGB images or depth images into the ZED SDK.
Stereolabs Support
Ok, thanks for your prompt responses. What do you recommend I do to reduce artifacts? The main issue is reflective/semi-transparent surfaces which ZED’s depth estimation don’t correctly take into account. I’m running NEURAL mode, high details, medium range. Is there any more customization that would reduce this problem?