Getting little to no data to mesh the model
Hello,
I’m by no means an expert, very intermediate still. I’ve used reality capture in the past with great success meshing models, but I have started working on a new project where I’m not having any success meshing models or getting anything close. In the past, I’ve shot objects outside or in well lit rooms with pretty good success. I was asked to help scan and mesh 20+ objects. I figured the best set up would be a black backdrop and put the items on a turntable. I set the turntable to turn around 5 degrees with over 60 pictures per full rotation. For each object, I would get a low angle, straight at the object angle and high angle to make sure I got great overlap. Each getting full rotations and maybe a few extra shots here in there. Totaling about 180+ photos per object.
The Camera I used was a Canon R5 that’s 45 megapixels. I shot in raw that I exported out of Lightroom as Tiffs. Most of the pictures were shot with an ISO below 400. I shot on a tripod and the shutter was 1/125 and the F stop was 8. The lens I used was a 35mm that I had a polarizer filter on and I used a flash to help with the lighting. In Lightroom I dropped the highlights and boosted the shadows. Other than that, I added a little clarity. In RC some of the models come in as components but even after adding control points they still won’t realign and mesh. When I start placing control points even the points RC suggests are pretty off. I’m also getting some of the objects that will come in as one component, recognizing most of the photos, but there is little to no data to mesh.
Happy to provide links to my files and photo sets of the issues I’m having. The screen shot below is of one of the models with the setting I used for alignment. Any and all help would be much appreciated. Is there any settings I’m missing or something I’m doing wrong? I think I got too over-confident with the success I’ve had in the past using RC and creating models, but now I’ve run into a situation I’ve never experienced. Thanks again!
-
Hello,
I would say that the issue is the background since RealityCapture is not singling out the object of interest, but the image as a whole.There are a lot of creases and shadows which may be recognized as features on multiple images, and therefore mess up the alignment. I would recommend trying to change the background to be as neutral (single color, no shadows, no distinguishable details) as it can be, or maybe using a mask layer to separate the object from interest from the background.
Masks can be created in a third party software and imported into RealityCapture:
Please sign in to leave a comment.
Comments
1 comment