FMRConv: Multi-Centroid Based Local Shape Perception for Deep Learning on Point Cloud

FMRConv is a spatial convolution operator for point cloud

This is an implementation of FMRConv by PyTorch.

Introduction

This project propose a new scheme, called Frame Multi-Relationship Convolution (FMRConv), for performing the 3D point cloud convolution and extracting the features from the individual cloud points.

Environment

This project passed the test in the following environment

Software

• PyTorch >= 1.12
• PyTorch3D >= 0.6.2 how to install
• NVIDIA CUDA Toolkit >= 11.5
• NVIDIA cuDNN >= 7.6

Harware

• NVIDIA TITAN RTX / NVIDIA Tesla V100 / NVIDIA GeForce RTX3090
• 32GB RAM

Classification

Data Preparation

Download alignment ModelNet here and save in data/modelnet40_normal_resampled/.

Run

python train_cls.py --log_dir [your log dir]

We provide a pre-trained model of FMRConv(Single Scale Grouping) here with an accuracy of 93.35%.

Part Segmentation

Data Preparation

Download alignment ShapeNet here and save in data/shapenetcore_partanno_segmentation_benchmark_v0_normal/.

Run

python train_partseg.py --normal --log_dir [your log dir]

Semantic Segmentation

Data Preparation

Download 3D indoor parsing dataset (S3DIS) here and save in data/Stanford3dDataset_v1.2_Aligned_Version/.

cd data_utils
python collect_indoor3d_data.py

Processed data will save in data/stanford_indoor3d/.

Run

python train_semseg.py --log_dir [your log dir]
python test_semseg.py --log_dir [your log dir] --test_area 5 --visual

Experiment Reference

This implementation of experiment is heavily reference to yanx27/Pointnet_Pointnet2_pytorch
Thanks very much !