Why is it -i images 2, can be removed? What's the impact?

[chenmaolin1]

mipnerf360 indoor

python train.py -s -m -i images_2 --eval --imp_metric indoor

Why is it -i images 2, can be removed? What's the impact? thanks

[fatPeter]

--images / -i is set for a consistent comparison with previous works (e.g., mipnerf, 3dgs). This setting is for research purpose only. Details can be found in https://github.com/graphdeco-inria/gaussian-splatting/blob/b17ded92b56ba02b6b7eaba2e66a2b0510f27764/full_eval.py

Note that some works use -r 2 instead of -i images_2, which is an unfair comparsion to some extent.

[chenmaolin1]

--images / -i is set for a consistent comparison with previous works (e.g., mipnerf, 3dgs). This setting is for research purpose only. Details can be found in https://github.com/graphdeco-inria/gaussian-splatting/blob/b17ded92b56ba02b6b7eaba2e66a2b0510f27764/full_eval.py

Note that some works use -r 2 instead of -i images_2, which is an unfair comparsion to some extent.
Thank you for your answer. So do I have to add -r 2 during training?

[chenmaolin1]

--images / -i is set for a consistent comparison with previous works (e.g., mipnerf, 3dgs). This setting is for research purpose only. Details can be found in https://github.com/graphdeco-inria/gaussian-splatting/blob/b17ded92b56ba02b6b7eaba2e66a2b0510f27764/full_eval.py
Note that some works use -r 2 instead of -i images_2, which is an unfair comparsion to some extent.
Thank you for your answer. So do I have to add -r 2 during training?
And what is the relationship between -r 2 and -i images 2 during training?

[fatPeter]

--resolution / -r indicates 'Specifies resolution of the loaded images before training'.

More details can be found in readme of the 3dgs repo (https://github.com/graphdeco-inria/gaussian-splatting?tab=readme-ov-file#running).