Internal Error MSS_STR001 . Why is this happening?

General Feedback & Requests
,
1

Hello there,

I have an issue with this error popping up on a regular.

Aligning Images
Unexpected program state
Internal error MSS_STR001

Why is this happening? I am attempting to align images taken from a helicopter rig. We had a rig of 4 overlapping RED cameras shooting at 60 FPS in 8K. The material looks great and I was previously able to align a shot with 2000 images (500 frames x 4 cameras) without a hitch. Since I convert the RED footage into JPG’s in another software, I don’t have any exif data to feed to RC, so I at least am giving it the 35mm equivalent of the focal length used. Still, while one set of images, using all of the 60 FPS worked well and produced a wonderfully detailed scan, other takes, that where flown in similar heights and with similar speeds constantly spit out this error. What am I doing wrong? What can I change in RC’s settings to get these images aligned. It is starting to get frustrating.

Cheers

Guillaume Wyatt

2

Hi Guillaume,
is there a problem with just one part of your images? Can you try keep empty information about focal length?
Have you changed the alignment settings? Can you please check if some of the images are not corrupted? IS it possible to process the cameras separately?

Please try to clear the cache, reinstall RC (you can launch RC with holding shift key and choose Make it like a clean install) and try again.

Are you using control points and how many of those do you have? Did RC crash, or only an error showed up?

Where are your images saved? Is this the same when you will use RealityCapture to create JPGs?

3
  1. is there a problem with just one part of your images?

No all images are fine. Loaded them in Nuke and rendered a movie of them without an issue.

  1. Can you try keep empty information about focal length?

I tried with and without entering a focal length. Both failed.

  1. Have you changed the alignment settings?

I had the alignment settings on default

  1. Can you please check if some of the images are not corrupted?

None of the images are corrupted.

  1. IS it possible to process the cameras separately?

It is, but then I would need to merge the components and I would rather have them align together as in the other shot I tried, that worked just fine.

  1. Please try to clear the cache,

Did that. Didn’t help

  1. reinstall RC (you can launch RC with holding shift key and choose Make it like a clean install) and try again.

Cannot do that at the moment, because I don’t have admin rights on my computer at work, meaning I cannot uninstall or install any software.

  1. Are you using control points and how many of those do you have?

No control points are being used at this point.

  1. Did RC crash, or only an error showed up?

RC never crashed. The error message simply appeared at the end of the alignment process.

  1. Where are your images saved?

They are saved on a local nvme ssd drive, so read speed shouldn’t be an issue.

  1. Is this the same when you will use RealityCapture to create JPGs?

I cannot use RC to create the JPGs since RC doesn’t know the format RED cameras write to (.R3D)

I will simply try other settings and hopefully find some that work I guess. Still thank you for your input.

Cheers,

Guillaume

4

Hi,
if the component workflow works, could you try align the cameras separately, export components and then import components to one new project and align?
For Make it like a clean install (opening RC with pushed SHIFT) you don’t need to have rights, you just need to know the login details. This will reset the RealityCapture to its original stage and it helps in some unexpected issues.

5

Thanks again for your help. I have now managed to open RC like clean install and now it seems to work. I will make sure to keep that in mind, when ever I get that error again.

Cheers,

Guillaume

6

I’ts me again. Even though it worked once, I am now back at getting this error constantly and this time there is nothing from the above fixes, that will fix it. It is absolutely random. I even tried to devide tha whole data set into smaller chunks, but even 200 images will spit out that error and before I was able to align up to 4000 images of the same size.

Component workflow also doesn’t work for me. When I try to merge multiple components, the app will simply stay at 0% processing bar and not moving for hours.

I seriously need help.

Cheers

Guillaume

7

Hi Guillaume,
have you changed something in your data? Are these the new data or the old one?
There has to be something wrong with your images. What is the difference between set which worked and the other one? Do you have 4 components for component workflow? Are you trying to align all of them in one project? If you will start alignment, is there something written in the console view?
Can you try to use other image format? How did you divide the data?
Would it be possible to share your data with us?

8

  1. have you changed something in your data?
    This is a new data set, but captured on the same day and shot using the same camera rig (custom rig with 4 RED 8K cameras)

  2. Are these the new data or the old one?
    This is the new data set.

  3. There has to be something wrong with your images. What is the difference between set which worked and the other one?
    Again, I don’t believe there is anything wrong with the images. I have checked all image sequences in a RAM player and also in a video editing program. If any file would be corrupt, then I would get errors in those other apps and I don’t. I exported the images as JPG’s from the original R3D Files from RED directly out of Nuke14 as 100% JPG. Since even with JPG’s the data set is 150 GB in total size (2050 frames x 4 cameras = 8200 images) I haven’t tried any other image format, because all others are a lot bigger in file size.

The images of course come without any EXIF data, because the meta data contained in the R3D files is for video purposes really and once exported from Nuke, the meta data of the JPG’s is empty. I have to enter the focal length in RC manually.

4.Do you have 4 components for component workflow?
Somehow I am able to align each cameras images by themselves so I end up with 4 components (CamA, CamB, CamC and CamD)
But other chunks of the same take, using the same camera rig, with identical camera offset to each other, I can align all images at once.

5.Are you trying to align all of them in one project?
Although I have successfully aligned another shot with 4000 images, I opted to try this one in parts, because I wasn’t getting anywhere. My plan was to align several parts of the shot and exporting .rcalign files and then importing those in a new project and merging the components, even though I did not have luck with that, because RC just hangs at 0% for ever just trying to merge two components, that each took about 10 minutes to align originally. So I haven’t gotten far with it.

This is the alignment I did before using 4000 images (1000 frames x 4 cameras). I aligned all 4000 images in one component and the camera path of the whole rig looks flawless.

ScanB_02
ScanB_021500×819 148 KB

  1. If you will start alignment, is there something written in the console view?
    I will copy and paste tomorrow what the console spits out, but basically I get the “Internal Error MSS_STR001” at the end of the alignment processing. In the case of not merging the components, the console just hangs for eternity on processing 0%. I have not waitet any longer than an hour today. The CPU was definitely doing something, but the progress bar wasn’t showing anything and after an hour I thought RC should have managed to merge something, given that the 4 components I tried to merge with enough control points, so that each component had at least 3 of them had aligned in about 10 minutes each.

  2. Can you try to use other image format?

I will try that tomorrow, but certainly not the whole 8200 frames as that would fill up my nvme ssd to the brim. At the moment I have less than a TB free on my ssd.

  1. How did you divide the data?

If you are asking about the chunks, I divided them into different folders. Part01 (first 250 frames x 4 cameras), Part02 (second 250 frames x 4 cameras). ect.

If you are asking about the generation of the images, then I loaded the original RED footage into Nuke14, and rendered the sequences for all 4 cameras out as a 100% JPG sequence.

  1. Would it be possible to share your data with us?
    I don’t believe that will be possible as it is for commercial purposes with different vendors involved. Sorry. Even posting the above screen capture might get me in trouble, even though it was a shot that got canned anyway.

As maybe a feature request, could you add verbose levels to the console’s output or is there the ability to write to a log file, that has a bit more info than just spitting out these cryptic error messages? If something is wrong with an image file for instance, it would be nice if RC would tell me exactly which one for instance. I am used to reading and understanding error logs on a daily basis when rendering scenes in 3d, So I would greatly appreciate if RC could write a log file for debugging purposes.

Just an idea.

Cheers

Guillaume

9

Hi Guillaume,
as I saw your alignment result it looks like the issue is in your images. There are just too many images that look the same and also the overlap is too high. Try to use every second or third image or create them in lower frame rate.

10

Then I don’t understand why it worked so well in one case where the circumstances where pretty much the same. Would be nice to have consistant behaviour. I tried the above example (the one with 4000 images) with less frames, meaning in one try I gave RC every fourth frame, so basically 1000 images and it failed to properly align them. I ended up with about 4 larger and 9 small components, but when I gave it all frames (60FPS) and all 4000 images, it was able to align all of them perfectly on the first try with only one component and with all 4000 images. So I guess it is just a matter of how moody RC is on any given day.
:roll_eyes:

11

Yes, but you wrote that the images were different in that case.
Is this dataset similar to one which worked?

12

Well it is different in that the terrain was a different one, shot on a different day, but with the same camera rig and technical setup.

The one that worked fine was more of a forest terrain with trees, bushes and grass, while the one that doesn’t work is a canyon with a dry riverbed and lots of rocks everywhere.

Other than that, same cameras. Same lenses. Same rig and therefore same position and orientation offsets. Same ISO. Same sharpness.

In both cases the Pilot wasn’t flying very fast, because we want to control the speed in editing by ramping fps up and down, so the motion blur is there, but not terrible.

I don’t know by which measures other than the ones I mentioned I can differentiate the two data sets. Maybe RC likes trees and bushes more than rocks.