is anyone else having issues with merging photogrammetry and laser scan components taking a long time? I’m currently trying to merge a large dataset consisting of 25k images and laser scans in 4 seperate components but using control points to bring them together. Back when I was using reality capture, this process would often take a day or two at the most, but it’s currently sitting on 6 days and only 7% complete. Is anyone else experiencing this?
Would someone be able to offer a faster workflow?
Hello Joel,
such long time is quite strange.
Is there something happening in the console view?
What were your repro steps and used alignment settings? Are you closing and re-opening the app between the components’ computation?
Thanks for the response.
Console view says that the project has finished feature detection and is currently matching. That seems to be what’s taking a long time.
I basically aligned the photos into usable components, then tried to merge them with LSPs. I am not closing and reopening the app.
There has been a change so its doing something, but very slowly. Currently it’s still Matching at 14% after 12 days. It looks like the current RAM utilisation of the machine caps at 43% and doesn’t go any higher. CPU usage seems to fluctuate between 50-60%
My PC specs are: Intel core I9, 192 Gb RAM, Nvidea RTX 4090 so it’s a solid machine.
I tried copying the project to another machine with similar specs, but I turned off the option for “using GPU to accelerate” before merging because I thought the VRAM of the GPU might be causing issues. That project reached 14% within a day, but now has since slowed down, and has only progressed 1% in a day.
The E57s were converted to LSPs on this currently RC version, and all the components I’m merging have been prealigned using high feature detection, and are currently set to merge using component features only.
As I wrote, this is quite unusual. It shouldn’t take 12 days just to align or merge the data…
Can you try to use Align option (as you wrote you used merging)?
Yea the align option does seem to work quicker, but my understanding is that using the align option risks breaking apart the registrations I’ve created.
I agree that it’s weird as it was never taking this long previously, but now we’re getting this same error consistently across a variety of different projects, on machines with different specs
Are the used control points placed well for all components? Which app’s version are you using? Can you try to close and reopen the app after the components’ creation? There is one bug with releasing a memory and for such big projects it can be an issue…
i have had the same problem!! I though it was just me and didn’t see anything on the forum. When i merge components using the component workflow it is also taking days long time. 2 of my projects have been merging since Friday, one is at 11 % and the other is now are at 37%, slowly moving but very slow. it can align 10,000 jpg images very quickly and with a better result than in reality capture but if i add LSP’s and align or merge it becomes so slow. defiantly something wrong when aligning to laser scans not just merging in the new version, i have tried both and both are taking a long time. i am using 2.0.1. i restarted reality scan between each alignment, i reset all setting, (i also tried turning off GPU for alignment and it goes to 10% quickly but then is also slow after) if you cancel alignment and then align again it sometimes goes faster
Can you check how many features was detected for the LSPs? There is a bug that more then set maximal value are detected and therefore it can take longer time in the merging process.
many of my LSP’s have detected many tie points up to 189k per LSP especially the cube face looking down or next to a wall. if the cube face looks to the side most are about 20-80k looking up at the sky has very few. my settings are set to default (max features 40k features per image 10k pre-selector) how can i make it select less features? my computer doesn’t looking like it has hit ram limits so is the number of feature is fine? it is only using 60 gb of 128 ram
It is too much and it can affect the computation (as there is a lot of features, all of them are checked with other features, which can multiply the computation time).
Unfortunately, there is no solution for now, but our developers are working on a fix for this.
thank you . if i align not merge will it speed it up or make it worse? and if i leave it and wait will it work but just slowly? mine is for a research project so as long as it works one day it is fine, it aligned the images better on the new version when i checked so if it works it is probably better for me. also if i turn off graphics card for alignment will it make a difference to make it faster or slower (i have a 5070ti)?
Yeah I might be able to share a dataset of around 10-12k inputs? Or is that too big? Whereabouts can I put the data?
The control points are placed well, we’re mostly using survey stickers. The latest version so 2.0.1. I close the app after component creation, then import the registrations into a new project.
Hi, according to tests there is no big difference between merging or aligning components. For some cases it is slightly quicker, for others it is slower. But the time is almost the same.
As we haven’t experience such behaviour, hard to say. But until you have free RAM, it should continue.
With turned of GPU in alignment it should be slower then.
There shouldn’t be a limit (or there is 500 GB per file). The mentioned limit is for your personal account. When you use the invitation link it should work.
Have you seen the window with question, if you are a part of Epic Games? If so, which option have you chosen?
Ohh ok. Thats good to know. Yeah I believe i said I was. Too easy. I’ll get started uploading those files tomorrow (about to finish up for the day here). Thanks Ondrej