View on GitHub


Created by Yangyan Li, Rui Bu, Mingchao Sun, and Baoquan Chen from Shandong University.


PointCNN is a simple and general framework for feature learning from point cloud, which refreshed five benchmark records in point cloud processing, including:

See our research paper on arXiv for more details.

Code Organization

The core X-Conv and PointCNN architecture are defined in ./pointcnn.py.

The network/training/data augmentation hyperparameters for classification tasks are defined in ./pointcnn_cls/*.py, for segmentation tasks are defined in ./pointcnn_seg/*.py


Commands for training and testing ModelNet40 classification:

cd data_conversions
python3 ./download_datasets.py -d modelnet
cd ../pointcnn_cls
./train_val_modelnet.sh -g 0 -x modelnet_x2_l4

Commands for training and testing ShapeNet Parts segmentation:

cd data_conversions
python3 ./download_datasets.py -d shapenet_partseg
cd ../pointcnn_seg
./train_val_shapenet.sh -g 0 -x shapenet_x8_2048_fps
./test_shapenet.sh -g 0 -x shapenet_x8_2048_fps -l ../../models/seg/pointcnn_seg_shapenet_x8_2048_fps_xxxx/ckpts/iter-xxxxx -r 10
cd ..
python3 ./evaluate_seg.py -g ../data/shapenet_partseg/test_label -p ../data/shapenet_partseg/test_data_pred_10

Other datasets can be processed in a similar way.