Implement the Point Net model for Classifications and Segmentation tasks.

Dataset: S3DIS dataset (http://buildingparser.stanford.edu/dataset.html#Download)

The original architecture is likely below the image:

s3dis

.
├── Area_1
│    ├── hallway_1
│    │    ├── color.npy
│    │    ├── coord.npy
│    │    ├── instance.npy
│    │    ├── normal.npy
│    │    └── segment.npy
│    ├── hallway_2
│    └── ......
├── Area_2
└── ......

In some cases, the .hdf5 type is preferred.

s3dis-hdf5 (use SaveAsNpyToHdf5.ipynb to convert)

.
├── Area_1
│    ├── hallway_1_partition1.hdf5
│    ├── hallway_1_partition2.hdf5
│    ......
│    ├── hallway_1_partition6.hdf5
│    ├── hallway_2_partition1.hdf5
│    └──......
├── Area_2
└──.......

The two main files are PointNetCls.ipynb and PointNetSeg.ipynb respectively which are for Classification and Segmentation tasks.

Enjoy and leave your feelings!