Changing the grid size resulted in incorrect inference results

[Forgeaheadye]

Once I change the ratio of point cloud to pillar size, the inference results are completely inaccurate.

const int grid_x_size = (max_x_range - min_x_range) / pillar_x_size;
const int grid_y_size = (max_y_range - min_y_range) / pillar_y_size;
const int grid_z_size = (max_z_range - min_z_range) / pillar_z_size;

If my own grid size does not match the grid size provided by the CUDA pointpillar author, the reasoning is incorrect.
This issue has caused me to be unable to freely set the point cloud range and pillar size.

[Forgeaheadye]

const float min_x_range = 0.0;
const float max_x_range = 69.12;
const float min_y_range = -39.68;
const float min_y_range =39.68;
const float min_z_range = -3.0;
const float max_z_range = 1.0;

const float pillar_x_size = 0.16;
const float pillar_y_size = 0.16
const float pillar_z_size = 4.0

This is the point cloud range and pillar size provided by the author.

Thank you very much for your answer

[bodehai]

I have the same problem. Have you finally resolved it? If it is resolved, could you please let me know where the problem is?

[brillint]

I have the same problem. Have you finally resolved it? If it is resolved, could you please let me know where the problem is?

[bodehai]

I won't modify the parameters of the model right now. Instead, I will crop the range of the point cloud in my own code and select the appropriate point cloud.   护玉者 ***@***.***  

…

------------------ 原始邮件 ------------------ 发件人: ***@***.***>; 发送时间: 2023年12月27日(星期三) 中午11:32 收件人: ***@***.***>; 抄送: ***@***.***>; ***@***.***>; 主题: Re: [NVIDIA-AI-IOT/CUDA-PointPillars] Changing the grid size resulted in incorrect inference results (Issue #98) I have the same problem. Have you finally resolved it? If it is resolved, could you please let me know where the problem is? — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: ***@***.***>

[brillint]

I won't modify the parameters of the model right now. Instead, I will crop the range of the point cloud in my own code and select the appropriate point cloud.   护玉者 @.***  
…
------------------ 原始邮件 ------------------ 发件人: @.>; 发送时间: 2023年12月27日(星期三) 中午11:32 收件人: @.>; 抄送: @.>; @.>; 主题: Re: [NVIDIA-AI-IOT/CUDA-PointPillars] Changing the grid size resulted in incorrect inference results (Issue #98) I have the same problem. Have you finally resolved it? If it is resolved, could you please let me know where the problem is? — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>

好的，谢谢啦