Sem_part issues for own data preprocessing

charlesq34 / pointnet

PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation

Other

4.66k stars 1.44k forks source link

Sem_part issues for own data preprocessing #166

Open l53ma opened 5 years ago

l53ma commented 5 years ago

Hi, thanks for your great work, I'm a freshmen to the 3D point cloud.

want to use the my own dataset for the pointcloud sementic segmentation, for now, my dataset is .h5 format with XYZ and point label. For each .h5 file, the shape for input XYZ is (2048, 2048, 3), The shape for input label is (2048, 2048). I would like to know how can I split each .h5 file into blocks with area 1m × 1m? what will be the outputs, does it still .h5 files? Thank you for your assistance.

TianyangChen357 commented 5 years ago

Hi,

I suggest u follow the gen_indoor3d_h5.py (dir: ./sem_seg/), in which the normals of each point are generated within a 1m1m1m cubic block (or window). The codes of this step are in indoor3d_util.py (dir: ./sem_seg/).

chester256 commented 4 years ago

Hi,

I suggest u follow the gen_indoor3d_h5.py (dir: ./sem_seg/), in which the normals of each point are generated within a 1m_1m_1m cubic block (or window). The codes of this step are in indoor3d_util.py (dir: ./sem_seg/).

The block size seems to be 1m by 1m by default height. They did not sample blocks along z-axis.