Questions on virtual points

[sun-yue2002]

MSMDFusion/mmdet3d/models/detectors/MSMDFusion.py

Line 388 in 7b5b274

xyz_normalizer = torch.Tensor([13.5, 13.5, 2.0]).to(device)

Hi!
I have 2 questions:

1. I wonder why this code is needed.
2. Besides I noticed that the original vitual point has 15 features, what these features are?

Hope for your response.
Best

[SxJyJay]

1. The normalizer is used to control the range of point cloud coordinate values.
2. You can refer to MVP for the specific meaning of these 15 features.

[sun-yue2002]

1. The normalizer is used to control the range of point cloud coordinate values.
2. You can refer to MVP for the specific meaning of these 15 features.

Thanks a lot for your response. I will check MVP for the answer.

[Libraaer]

1. The normalizer is used to control the range of point cloud coordinate values.
2. You can refer to MVP for the specific meaning of these 15 features.

Thanks a lot for your response. I will check MVP for the answer.

Hello, have you trained this model? I'm using A800 here but still not enough to train. If you've trained, can you give me an answer to the reason for my lack of memory here, thank you