under import file formats there is explained, that it is possible to open colmap files in Reality capture. But how is it done? I tried the Colmap exported “cameras.txt”, “images.txt” and “points3D.txt” but i always get the error “The physical end of the tape has been reached”. which file do i need to open and how can i Fix the error? (The alignement of Reality capture has alignement problems because between the Drone footage and the ground footage is not enougth overlap. This is why i did it with Colmap)
Hi stiv2804,
do you have also database_intrinsics_v1_1 file there?
Hi orthan,
First of all, thanks for the quick response.
In the export folder von Colmap i only have the 3 txt files and a project.ini file. So there is no intrinsics file.
Hi, I checked the error and it seems more like a Windows error: The physical end of the tape has been reached - VOX
According to this your disc is full and you need to delete something. Is it possible that your cache is full?
Which of the TXT files are you opening?
Hi,
Sorry for the long waiting period, i rebuild the colmap project in case something would be corrupted other there, still same issue.
on C:/ (Windows installation and Userfiles) is 80GB left, on the cach disk is 300GB left and on the disc where the files are is 110GB left.
I tried all 3 of the txt files. I have the same error with all of them.
For reference the size of the files:
cameras.txt: 14KB
images.txt: 1067981KB
points3D.txt 109166KB
Is there probably somewhere an option to increase the log level?
In the log i get the same text:
Importing failed after 8.189 seconds.
Processing failed: The physical end of the tape has been reached…
[0x5072\0x4999\0x5034]
[0x5072\0x4999\0x5034]
Loading Project failed after 8.879 seconds.
How big model is it? Or how many images are used in that project? Is it possible to share the files with us?
currently i can’t upload the files because of size, but im working on it. I have 1093 images and 867487 Points
This problem still persists.
It is partially caused by spaces at the end of the lines, And by empty lines at the end of the script.
Further, I don’t know that RS2 ‘needs’ for the points3D.txt file to be compatible, there is no documentation.
I read:
If you get an error like this, it is caused by the format of the XGrids point3D.txt file.
But no further explanation is given, cannot find.
I have a SHARE C1 LiDAR scanner and made a pose2colmap.py script that converts scanner output to COLMAP format. I have tested the script for direct drag&drop input to Postshot and LichtFeld Studio, but RS doesn’t wanna see it. Only when you delete points3d.txt contents which is part of the parcel. Makes no sense, this needs some love. Can provide (working elsewhere) sample data if needed
Hi @AVsupport, some examples would be usefull. Please send me your email in a direct message and I’ll create a folder for you where you can upload the data. We’ll take a look.
Hi Jakub, this would be extremely helpful.
I’ve been working on this and struggling with this for quite a while now, Trying different options.
I have created a working script that can create working COLMAP folders from my SHARE C1 LiDAR scanner, those read fine drag & drop into Postshot and LichtFeld Studio, but cannot be imported into RS2 sadly. If I’d know what the limitations are (besides the current syntax error, ‘end of tape’,) then i could maybe find a way.
I have uploaded these files already on my Google drive
These contain original (fisheye and undistorted) image sets, LiDAR , trajectory, intrinsics, also my pose2colmap.py script outputs (confirmed working) for LFS and PS (identical), with both versions pinhole and fisheye. These also contain masks.
Would be helpful if you could grab the data from there by upload is a little slow.
Hope this helps!
1 Like
Update: I have now created a version of the script that will allow for import into RS2.
Output files are in the google drive folder under COLMAP_RS2_fisheye
These now will load into RS2 (with local euclidian) and when point cloud is .las added, all in the same spot/orientation.
I have generated associated image masks in the folder, but I cannot see them being loaded into RS2 as expected.
What do I need to do to make this happen @JakubVanko ?
These are 120deg fisheye which should still be fine, apparently, but outside of the frame is black, so cannot be used without a corresponding mask.
Update: I have masks with correct file name, specs and folder structure. Tested as these will load fine with manual drag&drop image folder import into RS2. However, they are not recognized during COLMAP import.
Have tried to add those masks into cameras.txt and images.txt file, but RS2 will not reconigze those on import either.
discord bot Discord suggests CLI mask import after the fact is also not an option in RS2 (Was apparently working in RC).
There is also no option to import masks via the GUI after the fact, only Ai masking tools are offered.
How can I use existing external masks with COLMAP import?
Hi @AVsupport , we are checking the files.
Regarding the masks. The mask import with COLMAP is not implemented in RS. I’ll add it as a feature request.
There isn’t a GUI way of importing the masks (That is an existing feature request), but it is possible to import the masks after the COLMAP import as an image layer. Learn more about image layers here:
TLDR: The way that works for COLMAP is if you have the masks stored in the same folder as source images with the _mask suffix. See the example that worked for me:
Thanks Jakub, will give this a try!