Some advice for Newbies....
I am very new to Reality Capture but in a week or so I have gotten more and more confident with the program and wanted to share my big learns during the last week or so...
- Take, take, take, take and TAKE pictures, just keep on taking every possible angle. Think of it as you have a sphere or dome with your subject in the very centre and the shell of the dome or sphere representing where your camera has been. If you surround your subject with pictures and make small movements between each frame you should have no trouble aligning your 3D Scan. This advice is in so many tutorials but it is just so important. Technically it feels like if you take enough images you won't have to worry about control points and all that painstaking stuff!
- Try and keep exposure, white balance all even with nice sharp pics and don't be afraid to use f/11 for depth of field.
- If a part of your model has holes or just isn't rendering right, if it hasn't moved and you can, go back out and take more more detailed images of that part (and lots of them) then come back and try and realign and those new images should just slot right in and your holes hopefully will disappear.
- Really smooth or shiny surfaces will give you problems.
- Once you have alignment sorted the rest is quite straightforward but the most important part of the process is alignment. Get that right and your model will be great! If you align and you have multiple components in Alignment settings try changing overlap to Low and Increase features per image I often use 80000. In the advanced settings try increasing preselector features to half the Features per Image number so 40000 and Increase Detector Sensitivity to High (be careful if using ultra - long process times and could get weird effects)
- Watch all the tutorials you can on Ground Point and Reconstruction Boxes get those two sorted and life is so much easier!
That's my take on Reality Capture after almost two or so weeks. I made a model of some bananas and a pineapple on our kitchen island the other night, I was amazed at the detail RC recreated, even all those spiky bits that stick out from the pineapple were rendered perfectly in 3D.
Good luck with your learning the software. I'm now testing drones and having some great success in making 3D scans from the air.
Cheers
-
This something I'm yet to try but doesn't this turn the dome/sphere principle on it's head? Rather than the camera being the shell of the sphere it's now right in the middle and what's more aren't you having to effectively capture 100's of spheres/domes of image data to capture the inside of a room? That does sound quite demanding and no wonder you are having issues. From what I have read are blank walls also a big issue unless you have laser scanning? Still haven't figured out how you cold laser scan to get the geometry of a blank wall then use images to fill in the texture as unless you took a photo from every laser scan point how could you then tell the software the imagery of the walls if it can't align wall shots?
One last question, in having no success in multi-room building interiors is that all softwares or just RC?
Cheers
Please sign in to leave a comment.
Comments
2 comments