MeshroomCL 0.7.3 error on MultiveiwStereoCL node

[JesseSchilbe]

I am using meshroomCL 0.7.3 on my bootcamp iMac running window 10 pro. I was able to make two 3d scanned model no problem but when I try with 3 other image sets It stopped on the MultiveiwStereoCL. this node is not listed in these two links Manual: https://bit.ly/3jLrx8p Deep Dive of Nodes: https://bit.ly/3teSzbm . please help. The link to all the photo I used are link here https://bit.ly/3zM7ook the password to access the link is: makertales the photos that I have problems with are the ones with the names Hare, Owl, Snail, Sandstore. I made a folder with my logs in it https://drive.google.com/drive/folders/1llnAq_-1m6CReU4CLHgpXWcHFq-4Dug3?usp=sharing

[revisionarian]

Hi @JesseSchilbe, thanks for posting your bug report and data; sorry that you have encountered some problems with MeshroomCL.

There are two separate problems occurring here:

1) On the Hare, Owl, and Snail datasets, you are running out of video memory (just barely!). Do you know how much VRAM you have available on your graphics card? To work around this problem, you can change the "Max Image Size" slider in the MultiviewStereoCL node attributes from the default value of 3200 down to something smaller...reducing it to 1600 would use 4x less VRAM, for example. Also, close any other graphics programs that might be running on your computer at the same time that MeshroomCL is running, so that the full video memory is available for MeshroomCL.

2) On the Sandstone dataset, the failure is actually happening in the image alignment step (the StructureFromMotionCL node). Only two of the cameras are getting aligned, so the MultiviewStereoCL node then fails. To work around this problem, you can download and run the latest MeshroomCL version 0.8.0, and then switch to the "Photogrammetry (OpenCL MVS)" pipeline under the File → New Pipeline menu before importing your images. This pipeline uses a different image alignment algorithm that works better on the Sandstone dataset.

Please let us know if these two workarounds address your problems. In the meantime, we will do some more debugging analysis on the datasets you provided so that MeshroomCL will work better with them "out of the box."

[JesseSchilbe]

I found info on my graphics card and video memory on my bootcamp. Name: AMD Radeon Pro 5300 Total Available Graphics Memory: 8115 MB Dedicated Video Memory: 4096 MB System Video Memory: 0 MB Shared System Memory: 4019

[revisionarian]

Ok, so you have a 4 GB video card...I would have thought that is plenty enough for the MultiviewStereoCL node to operate at 3200 pixel resolution, but your error log shows you are just slightly exceeding the available memory.

I think you just need to reduce the "Max Image Size" slider in the MultiviewStereoCL node to something in the 2400-3000 pixel range, and then it should work within the available memory.

[revisionarian]

I mis-diagnosed this issue on the Hare, Owl, and Snail datasets as running out of video memory. Instead, I now see that the problem is running out of RAM. How much RAM do you have on your computer?

You can still workaround this problem the same way, by changing the "Max Image Size" slider from 3200 to a smaller value. We will release a new version of MeshroomCL soon that fixes this problem by using less RAM during processing of the MultiviewStereoCL node.

[JesseSchilbe]

I downloaded MeshroomCL 0.8.0 and used the new pipeline on the Sandstone image set and it worked. Next I tried the Hare, Snail, Owl image set and this time the red bar appeared on the texturingCL node. I made sure to change the max image size to a lower number I changed it to 1400 and did the same in StructureFromMotionCL. I also changed the max CPU threads StructureFromMotionCL to 7. Next I tried a image set called Ammonite that worked for me fine in 0.7.0 but in did not in 0.8.0 I did not have any redbars but it only saws the front of the model .

I made a google dive folder of the logs and other stuff but its not done uploading yet but I will post a link when its done .

[JesseSchilbe]

8 GB of ram on my computer by the way

[JesseSchilbe]

Here are the links to my logs. I was going make everything able but I figured it will be easier just to make the logs able. https://drive.google.com/drive/folders/1QL5T0BHYO4a6QLH5ScNR_sComkpje2rV?usp=sharing https://drive.google.com/drive/folders/1flO_-UOQsNgzzUz61e8Nna2INaOyGbgL?usp=sharing https://drive.google.com/drive/folders/1swfesbgNp_kTOcXtZC7QZdq1IOnYUqnJ?usp=sharing https://drive.google.com/drive/folders/1mqZFuPAJKP_6ag-A-zjGV3PTeduwrHAS?usp=sharing