convert.py resizing images without resize flag

[Mikey-E]

I have a custom dataset that I have been resizing with convert.py from https://github.com/graphdeco-inria/gaussian-splatting. The images folder it creates (from "input") is of different size than input. e.g. input will be 666M in size but images will be 418M. I am able to run gaussian models with either folder set as "images" but I would like to know - is this change in size after convert.py is run something to be expected or not?

[niujinshuchong]

Hi, the script will resize the image if you enable the resizing. It is fine to use different size of images for training. Smaller images trains faster.