Open reconlabs-sergio opened 5 months ago
The simplest way to ensure the image size divisible by 64 is to set h=h//64*64, w=w//64*64
.
If you set different max_h
, max_w
, these codes would change the intrinsic automatically:
https://github.com/maybeLx/MVSFormerPlusPlus/blob/5f9264ed98a4376c91fcdfd7562f8cdc3014647d/datasets/general_eval.py#L120
Hi,
In your readme, it's said that:
"the resolution of input images must be divisible by 64" My original images have 3840*2160 resolution but colmap creates an
images
folder with size 2127x3779.Is it okay to resize the image to a size that's divisible by 64, like 3840*2176 ? Would it be necessary to change the camera parameters if one resizes the images?
What would be the correct way to process a dataset where images are not originally multiply of 64?