Hi,
We are currently using a ZED 2i with the PERFORMANCE depth mode, but we understand it has been deprecated since SDK 5.0 in favor of the new NEURAL-based modes (NEURAL, NEURAL_LIGHT, NEURAL_PLUS).
We are looking for a non-AI depth mode, i.e one based on classical stereo matching (like the old PERFORMANCE, QUALITY, or ULTRA modes) rather than neural network inference.
Is there any plan to maintain or reintroduce a classical stereo matching depth mode alongside the AI-powered ones?
Thanks!
Hi
While deprected, they are still available for all ZED SDK 5.x. You can still use the enum in the API to enable them.
They could be removed in a future major update but it’s not currently planned
I don’t get it. Getting depth without AI mode on a stereo camera seems like an obvious capability, regardless of the release version.
How and why was the decision made to remove such a fundamental feature from a stereo camera? Am I missing something?
The whole point of using a stereo camera is to get depth without consuming too much GPU resources. Removing that basically opens the door to migrating to a monocular camera instead.
Computer vision depth modes are available for all released ZED SDK versions and are not currently planned for removal.
AI modes are order of magnitude more accurate and NEURAL LIGHT can still be faster than ULTRA.