Hey all!
There’s something I’ve been trying to figure out but it still isn’t working for me:
I have a rig with several cameras at certain positions that don’t move. I can scan an object with a lot of texture and it works just fine but if the object has less texture, Reality Capture often has trouble positioning some of the cameras in the alignment process (it’ll maybe get half right and the other half wrong).
I know that I can manually go in and set control points, etc. but I also know… or… I think I know… Isn’t there a way to SAVE rough camera positions from one scan (with an object that has a lot of texture) and then tell RC to use those same camera positions for all my other scans? What is this called? Or what is the process to achieve this?
I’ve tried a few things but so far, nothing has worked. If anyone could help point me in the right direction, tell me what this process is called, give tips or instructions, etc… I’d be very grateful!
Have a look at the XMP workflow (reuse camera alignments), covered in the doc and on the forum.
Hey Jonathan,
Thanks for the response! I have checked into the CMP workflow but the problem I’m seeing is: That seems to be for rigs that have a projection system maybe? Where one set of images has a model that has light patterns projected onto it and one set that does not have light patterns projected onto it. As in, it’s using the exact same model just with different texture patterns.
What I’m doing, since I don’t have a projection system in place, is placing a highly textured object in the center of the rig, capturing a set of images to get camera locations, and then placing numerous other objects that need to be scanned into the rig. What I was hoping was that there was a way to completely lock the camera locations so that no matter what shape or size of object goes into the rig, it’ll know where the cameras are.
Or does the XMP workflow work in this way?? It seems like with the XMP workflow, if I use a source folder with a new set of images but keep the old XMP files (from the highly textured model) and then hit “Align” - it seems to just completely ignore those “locked” camera locations and tries to resolve where the cameras are located based on the textures in the new image set… instead of using the camera positions from the last scan.
I keep searching for a solution but so far… no luck. Or I keep using the XMP workflow incorrectly. I’m not sure.
XMP workflow works for any workflow where you have repeating alignments with the same camera positions. So yes, a good way to work is align with any highly featured object, and then keep the xmp files. This is just a matter of renaming xmp files to new names to link to new pictures. Try to follow the documentation, there is well explained use cases.
You need to put the xmp files besides the source images and RC will find them.
The topic in the doc is named “Re-use Alignment”.
Jonathan_Tanant,
Thanks so much for the responses! Is there a “document” somewhere that I just don’t know about? I’ve only found stuff through the forum here, support articles and youtube videos… didn’t know there was a “document.” Could you point me to that? Because I’ve done several google searches for it now but still haven’t found it :/
You have to open a “Help” tab in Reality Capture (the small label tab where you can choose from “3d”, “2d”, “Console”…).
And then you can choose from topics. Helps is kind of dynamics, when you do some operations, the help will update itself and point to the relevant page, which is not so usual but very useful.
Jonathan_Tanant,
You’re my hero… this is exactly the type of tutorial I was looking for! Looked everywhere except for IN the software. Strange that they don’t just have some sort of user manual PDF or something kept outside of the software. Anyway, thanks a ton! I’ll give this a shot and see if I can get it to work. It looks like it’s exactly what I was trying before but I assume I missed an essential step or something.