Many times lately, I’ve had RC seem to just “give up” during alignment. Here’s how it was this morning after leaving it to run last night. As you can see in the screenshot, It’s stuck at 40% on “Processing.” It’s not using the CPU or reading / writing to the hard drive. It predicts that it will finish at about 3am, which is 5.5 hours in the past (see clock at bottom right of screenshot).
When it gets like this, the “abort” button does not work. If you press it, you will get the “aborting process…” screen but that screen will stay there forever. Force quit (“end task”) is the only way to close the program, and it is impossible to save your work.
I suspect this is the result of asking RC to do an impossible alignment. I realize that fixing this bug may be equivalent to solving the halting problem! However, at least the ability to abort the process and save work should be maintained. Also, it seems like there might be something really wrong here because the CPU use stops.
this looks like an issue inside alignment. Alignment of 1300 images should not take longer than 2 hours. Especially on the computer you have
If you find a data-set with less then 1300 images, e.g. 100 where you can repro the issue than it would help to find the problem.
Nevertheless we will be releasing a new alignment engine which should not have this problem.
thanks for posting.
Hi Admin, I’ve been trying to reproduce this in a smaller, simpler project, but so far I can’t. It only seems to happen on my 1300 image project with multiple pose groups. I ended up creating a new project anyway and it’s working fine. I’ll post here if it ever happens again.
It is happening to me as well on a 2500 img project.
When are you planning to release the new version with the improved alignment engine?
Can we do something meanwhile?
As for the alignment part, RC works well even with 25k image datasets, and issues like this one is mostly due to an inappropriate capture scenario, like cameras from almost the same position (small movement among poses) or “panorama capture”.
Can you send me your dataset that is not aligning correctly to email@example.com?
Thank you Wishgranter,
The overlap between images is strictly 90% and I captured the images following a parallel movement “mosaic” style. Just like in UAV but at a close distance.
The data set is huge, I don’t think I can send it by email
(There is) no need to send it per email, you can use Dropbox, Wetransfer or other data sharing services
Is it possible to send me, say, 50-100 images only to see what are you working with?
What alignment settings did you use there? Can you make a screenshot?
Yeah its even huge for we transfer hehe, the whole data set is around 400gb.
I think I can send 100 images, there are some copyright issues that I should ask first though.
Oh sorry, the full post is on “General” tab, a friendly forum user pointed me to this post.
This is my post:
Intel Xeon CPU E5 2687W v3 3.10 GHZ
GPU Titan GTX
Number of images: 2.500
MPX: 50 per image (Canon 5DSR)
Files: Jpge files
The process is stuck infinitely at the “processing” alignment step 66%
Relevant Alignment settings:
Max features per mpx: 1.000
Max features per image: 800.000 (With 50mpx its recognising around 574.000 points per image) (Is it completely useless to use that amount of points?)
Image downscale factor: 1
Max feature reprojection error: 1.5000 (Maybe I should lower this value down with the amount of points that are being recognized?)
Detector sensitivity: Ultra
Image overlap: High
PD: My idea is to achieve the highest quality results as possible.
Is there any reason on why RC is not using CPU extensively on depthmap calculation ? It just happened to me on this project apparently, I remember it being 97% on others?
CPU is NOT used in actual Depth map calculations, the CPU just “prepare data” for GPU to be processed.
Ah true! my mistake :lol: actually I had that written down in a document compiling info. GPU usage is about 30%, is that intended?
Yes that is OK , that are the CUDA drivers compiling and sending data to GPU for actual processing…