Hi,
I tried to construct a large mesh model using RealityCapture, but even though it finished successfully, some parts of the mesh could not be built.
During alignment, the camera pose was correctly estimated for those parts.
I was wondering, is there an upper limit to the number of images required to successfully build a mesh?
By the way, the mesh quality is “normal”.
Below are the settings at that time.
・Number of images: 7600
・Maximum number of feature points during alignment: 160000
・Mesh quality: normal
Hi,
the maximum number of processed images depends on your machine’s parameters.
Which version of RealityCapture are you using? What are your reconstruction settings? Do you have 2 GPUs? When you click on the model in 1Ds view, do you see Continue button?
Just a note, it is not necessary to use such big Max number of features. 80 000 should be the upper border (as you can run out of your RAM with such setting).
Here are my Desktop PC specs and the answer:
・CPU : Ryzen ThreadRipper 7970X
・GPU : NVIDIA GeForce RTX 4090（1 GPU）
・RAM : 256GB
・RealityCapture version : 1.4
When you click on the model in 1Ds view, do you see Continue button?
→ No…
Currently, I have set the Max number of features back to 80,000.
I tried again and was able to create a complete mesh, but for reference, please tell me how many images are needed to create a mesh.