How to predict on custom point cloud without labelling?

[tianyufang1958]

Hi,
I have successfully trained and test the model with our custom dataset, but I am wondering how I can inference the model on new point cloud without annotation? Do you have code for this?

I tried pcd_demo.py in mmdetection3d, but got the following error:
AttributeError: 'NoneType' object has no attribute 'dataset_meta'

[filaPro]

Unfortunately we don't have this code, please have a look at other issues e.g. #9, #57.

[tianyufang1958]

Unfortunately we don't have this code, please have a look at other issues e.g. #9, #57.

I can save the resulting point cloud without problem based on #57, but I would like to use the code in practice for input point cloud without annotation. Have you got any plan to work on this?

[GH-Fonic]

Does anyone use unlabeled point clouds to output prediction results?

[tianyufang1958]

Does anyone use unlabeled point clouds to output prediction results?

I am wondering the same thing.