Skip to content

Conversation

@rpavlik
Copy link

@rpavlik rpavlik commented Jul 6, 2021

Builds on #48 but this one is not complete - it looked like the CUDA dep was much deeper in the app, sadly, since that's what I wanted to try out first. I probably won't have time to do more with this for a while, but feel free to take this as a starting point.

@rpavlik
Copy link
Author

rpavlik commented Jul 6, 2021

Actually, it looks mostly like the main blocker is a pure-cpu version of the PixelCornerProjector.

@rpavlik rpavlik force-pushed the optional-cuda-app branch from a0609ca to c800636 Compare July 6, 2021 16:39
@puzzlepaint
Copy link
Owner

Regarding PixelCornerProjector: I think that this should only be used by the stereo depth estimation tool, which is not required for the calibration process itself, and (due to lacking optimizations) is already very slow when running on a GPU. So this tool should be even less practical on a CPU in this form. Thus, in case your use-case does not require it, perhaps the depth estimation tool could simply be #ifdef'd out for now if CUDA is not available.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants