About how to generate colmap points by the lidar.

[jiangxb98]

Can the author provide the result of processing lidar into colmap data? The index of the 3DGS run after our processing is 25.xx, which is not up to the results in the paper. How did you handle it? Is it possible to make the processed data public?

[qiuqc1]

Can the author provide the result of processing lidar into colmap data? The index of the 3DGS run after our processing is 25.xx, which is not up to the results in the paper. How did you handle it? Is it possible to make the processed data public?

Hello, can I ask how you got it working? When I tried, I found that I had to change a lot of code, and the lidar prior ply didn't seem to be given. Could you give me some guidance?
您好，能问下你们是怎么跑通的么，我尝试的时候，发现要改许多代码，而且lidar的先验ply貌似也没有给出，能否麻烦您给个指导呢

[jiangxb98]

Can the author provide the result of processing lidar into colmap data? The index of the 3DGS run after our processing is 25.xx, which is not up to the results in the paper. How did you handle it? Is it possible to make the processed data public?

Hello, can I ask how you got it working? When I tried, I found that I had to change a lot of code, and the lidar prior ply didn't seem to be given. Could you give me some guidance? 您好，能问下你们是怎么跑通的么，我尝试的时候，发现要改许多代码，而且lidar的先验ply貌似也没有给出，能否麻烦您给个指导呢

The code given by the authors is based on Deformable 3DGS (not consistent with the paper), it is not 3DGS, so we are reproducing 3DGS ourselves. The process is to overlay the lidar points of the whole scene (both key and non-key frames can be tried), and the color of the point cloud is also important, which can be directly projected to the image.

[jjlinghu]

Hi,

Can the author provide the result of processing lidar into colmap data? The index of the 3DGS run after our processing is 25.xx, which is not up to the results in the paper. How did you handle it? Is it possible to make the processed data public?

Hello, can I ask how you got it working? When I tried, I found that I had to change a lot of code, and the lidar prior ply didn't seem to be given. Could you give me some guidance? 您好，能问下你们是怎么跑通的么，我尝试的时候，发现要改许多代码，而且lidar的先验ply貌似也没有给出，能否麻烦您给个指导呢

The code given by the authors is based on Deformable 3DGS (not consistent with the paper), it is not 3DGS, so we are reproducing 3DGS ourselves. The process is to overlay the lidar points of the whole scene (both key and non-key frames can be tried), and the color of the point cloud is also important, which can be directly projected to the image.

Hi, do you mean that the released code of this paper is based on Deformable 3DGS? and the colmap data which is used to overlay the lidar points also can be produced in 3DGS as shown in there?

[jiangxb98]

Using only keyframe point clouds, point cloud colors are projected directly onto the image. Then write a data loading code for the nuscenes dataset. Information includes, internal and external parameters, point cloud location and color.

[jiangxb98]

I recommend taking a look at other open source 3dgs automatic driving scene reconstruction code to learn how to handle the data. This way you can get a clear picture of the data. And then you can go to process the nuscenes dataset. If you understand the data entry process, this code is not difficult to develop.

[jjlinghu]

Thanks for your advice! I will spend more time learning related codes.

[jjlinghu]

what a coincidence!

[wcyjerry]

I recommend taking a look at other open source 3dgs automatic driving scene reconstruction code to learn how to handle the data. This way you can get a clear picture of the data. And then you can go to process the nuscenes dataset. If you understand the data entry process, this code is not difficult to develop.

Is there some open source 3dgs automatic driving scene reconstruction code recommend, it would be of great help