Dear,
I'm not a CMU student, so I don't have access to piazza. I think this assignment is finished, so I raise my question here hoping to get a solution. I hope to achieve what is described in the task 1.3.2 Training On a Harder Scene. I followed the source code of gaussian splatting and made modifications. But the results I got were not very good. In the densification stage, good results cannot be obtained with the same threshold as the original text. I also tried increasing the number of initial points, which caused the video memory usage to easily exceed 20G. What measures should I take?
By debugging the source code of gaussian splatting, I found that its implementation can achieve high accuracy without any adjustments.