Depthmap with 2 different cameras

[Chris Arthur]

Just a general question concerning using multiple cameras:  How does this complicate the depth mapping node?  I have 2 cameras. With both of them separately I've had good results, but when I put the feed from two of them together, I'm finding that depth map either crashes with a red bar, or else I reduce the settings so low that it doesn't really benefit me at all to use two.  Are there settings that I should change on the cameras?  Does it matter if the image resolution is different?

Chris Arthur 

[Fabien Castan]

Hi,

I would be curious to look at your dataset.

Using multiple cameras should not create any problem for the depth map estimation.

You may have a misalignment in the SFM between the 2 which would explain a degradation.

But in general, with more images, you get more constraint and it should improve the final result.

Best,

Fabien

[Chris Arthur]

Actually, that was my feeling generally also:  more information means better models.   Through the debug on depth map, the difference in image resolution seemed to frustrate the choice of tile sizes.  Yesterday I tried a second time with the same subject, and I adjusted the resolutions to be the same in both cameras.  This time it worked well and got a decent result without any complaints.  So far, the process is to record 2 videos from the 2 cameras as they move around the subject and then to use ffmpeg to get a number of frames from the videos.

--
You received this message because you are subscribed to the Google Groups "AliceVision" group.
To unsubscribe from this group and stop receiving emails from it, send an email to alicevision...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/alicevision/6dbbd419-c380-4055-9608-589196c32874n%40googlegroups.com.