KangPeiLun

icon indicating a website link www.github.com/kangpeilun icon indicating an email [email protected]

icon company 东北林业大学 icon location 黑龙江 icon location Im reading a PhD, and Im happy to share my code with you during my researching. What's more, I'd like to discuss the issues with you, which will help all of us.

Results 10 comments of KangPeiLun

无法通道您的代码得到与您论文中相近的结果

我没有解决问题，丢在一边就没再管了

Memory error

> 你可以看我的仓库，我对图片裁剪进行了复现，裁剪后的图片存放在与原数据文件夹相同的子目录中。https://github.com/kangpeilun/utils-for-img-process/tree/main

Remove the need for manual Manhattan alignmentm, use COLMAP instead

Thank you. I tried Manhattan alignment using colmap before, but it was z-vertical instead of y-vertical as in the paper. I am not very familiar with the use of colmap,...

python train_vast.py : ValueError: min() arg is an empty sequence

Which dataset did you use for the test? I have not encountered such problems with `tandt/train` data

ValueError: need at least one array to concatenate

You can use meshlab to check the tiling and note that it is a good idea to perform Manhattan alignment before training vastgs

AssertionError: would build wheel with unsupported tag ('cp37', 'cp38', 'linux_x86_64')

simple knn has not been modified in this implementation, so you should ask the author of 3dgs to solve this problem

results of your reproduce

This implementation may still have a few bugs, as I mentioned in the README, so I thought I'd optimize my code first and then experiment, it shouldn't take long

Visibility-based camera selection

Thank you. Let me check what you mentioned

Image file suffix case issue

This is a relatively minor issue, which I will try to fix later

Is JAX needed?

As you can see, JAX only in image_utils. Py were used for color correction, the realization of reference [multinerf](https://github.com/google-research/multinerf) so I don't have to make too many changes to the...