Wenxuan Wu, Qi Shan, Li Fuxin

This is the PyTorch implementation of our paper [PointConvFormer]

We introduce PointConvFormer, a novel building block for point cloud based deep neural network architectures. PointConvFormer combines ideas from point convolution, where filter weights are only based on relative position, and Transformers where the attention computation takes the features into account. In our proposed new operation, feature difference between points in the neighborhood serves as an indicator to re-weight the convolutional weights. Hence, we preserved some of the translation-invariance of the convolution operation whereas taken attention into account to choose the relevant points for convolution. We also explore multi-head mechanisms as well. To validate the effectiveness of PointConvFormer, we experiment on both semantic segmentation and scene flow estimation tasks on point clouds with multiple datasets including ScanNet, SemanticKitti, FlyingThings3D and KITTI. Our results show that %the deep network built with PointConvFormer substantially outperforms classic convolutions, regular transformers, and voxelized sparse convolution approaches with smaller, more computationally efficient networks.

1. We introduce PointConvFormer which modifies convolution by an attention weight computed from the differences of local neighbourhood features. We further extend the PointConvFormer with a multi-head mechanism.
2. We conduct thorough experiments on semantic segmentation tasks for both indoor and outdoor scenes, as well as scene flow estimation from 3D point clouds on multiple datasets. Extensive ablation studies are conducted to study the properties and design choice of PointConvFormer.

1. Install dependencies

pip install -U ipdb scikit-learn matplotlib open3d easydict
conda install pytorch==1.8.0 torchvision==0.9.0 torchaudio==0.8.0 cudatoolkit=10.1 -c pytorch
pip install tensorboard timm termcolor tensorboardX

2. Compile cpp_wrappers

cd cpp_wrappers/
sh compile_wrappers.sh
cd ..

1. Please download the ScanNetV2 dataset.

2. Split the dataset file with data_preparation/split_data_label_ply.py. Please change the split and data_path accordingly.

3. Preprocess the dataset with data_preperation/prepare_data.py. Please change the basepath accordingly.

1. Before training, please setup the train_data_path and val_data_path in configWenxuanPCFDDPL5WarmUP.yaml;

2. You might also want to set the model_name, experiment_dir accordingly in configWenxuanPCFDDPL5WarmUP.yaml;

3. If you use turibolt and would like to save the results to blobby, Please set the save_to_blobby to True, and use_tensorboard to False in configWenxuanPCFDDPL5WarmUP.yaml;

4. Change other settings in configWenxuanPCFDDPL5WarmUP.yaml based on your experiments;

5. To train a model with n gpus, change the num_gpus and devices_ids accordingly in configWenxuanPCFDDPL5WarmUP.yaml. Change the --nproc_per_node to be n in run_distributed.sh.

6. sh run_distributedsh

Please download the pretrain weights at here

Then, you can evaluate with the following comand:

python eval_ScanNet_PCF.py --config ./configWenxuanPCFDDPL5WarmUP.yaml --pretrain_path ./pretrain/[model_weights].pth --vote_num 1 --split validation