colmap
Reconstruction with 3 images and 900 points
I tried to use my own data to reconstruct a point cloud. It's an image set I got from a 360 video I took around an item. Then I used image matting and tried to get a better result.
But at the end of the process, colmap only used 3 images to reconstruct. Does anybody know why is this?
I use colmap 3.9.1 on Windows10.
It really bothers me.
Thanks!
How many images you get in dataset? My practice: Angles from image to image should be around 20 degree or less.. (my practice - so not writen in stone ;)) Do not remove background - this help colmap to catch very narrow/slim items
@jaco001 Thanks for your reply. I have around 100 pictures for my data. I fixed the problem by changing the camera to PINHOLE. You can check this website https://colmap.github.io/cameras.html. The reason why I removed the background is that I am testing a single object reconstruction.