How to use this repo for 3d Point cloud data classification?

[jumptiger66]

Hello, Thank u for your amazing work. I try to use this repo for 3d Point cloud data classification but fail . Perhaps i miss some important issues , have you successfully applied it to Point cloud data ? Best regards !

[jacobgil]

Hi, I did not yet, But if you can share pointers maybe I can help make it work. Are there any pretrained open source models and examples we can use?

[jumptiger66]

I recommend you use this repo https://github.com/yanx27/Pointnet_Pointnet2_pytorch. Firstly, model1 (https://github.com/yanx27/Pointnet_Pointnet2_pytorch/blob/master/models/pointnet_cls.py) is a simple MLP network and model2 (https://github.com/yanx27/Pointnet_Pointnet2_pytorch/blob/master/models/pointnet2_cls_ssg.py) is a general local relationship awareness network in point cloud. Secondly, you can use dataset in this repo named modelnet10 or modelnet40 , instance dimension in this dataset is N * 6，N is the number of points, 6 is the coordinate (x,y,z) and normal line.And , you can use code (https://github.com/yanx27/Pointnet_Pointnet2_pytorch/blob/master/data_utils/ModelNetDataLoader.py) to get dataloader .This is counterintuitive. There are only two dimensions of 3D point cloud data. Finally, you can use code (https://github.com/yanx27/Pointnet_Pointnet2_pytorch/blob/master/train_classification.py) to train these model. Thank u for your help !

[EricLina]

waiting for your update!