The full paper is available at: https://arxiv.org/abs/2003.05420
The code has been tested with Python 3.7 on Ubuntu 16.04.
- tensorflow 1.14
- h5py
- IPython
- scipy
Download 3D indoor parsing dataset (S3DIS Dataset). Version 1.2 of the dataset is used in this work.
python collect_indoor3d_data.py
python gen_h5.py
cd data && python generate_input_list.py
cd ..
- Compile TF Operators
Refer to PointNet++ - Training
python train.py
- Estimate_mean_ins_size
python estimate_mean_ins_size.py
- Test
python test.py
- Evaluation
python eval_iou_accuracy.py
This code largely benefits from following repositories: PointNet++, SGPN, DGCNN, DiscLoss-tf and ASIS