Computer Graphics Forum (Proc. SGP), 2023

This is the official implementation of Cross-ShapeNet, a deep learning method that propagates point-wise feature representations across shapes within a collection for the purpose of 3D part segmentation. For more technical details, please refer to:

Cross-Shape Attention for Part Segmentation of 3D Point Clouds

Marios Loizou, Siddhant Garg, Dmitry Petrov, Melinos Averkiou, Evangelos Kalogerakis

Project Page | Paper (arxiv, CGF) | SGP 2023 presentation

Left: Given an input shape collection, our method constructs a graph where each shape is represented as a node and edges indicate shape pairs that are deemed compatible for cross-shape feature propagation. Middle: Our network is designed to compute point-wise feature representations for a given shape (grey shape) by enabling interactions between its own point-wise features and those of other shapes using our cross-shape attention mechanism. Right: As a result, the point-wise features of the shape become more synchronized with ones of other relevant shapes leading to more accurate fine-grained segmentation.

Follow this guide for the MinkowskiNet experiments on the PartNet dataset.

To conduct the MID-FC experiments on the PartNet dataset, please follow the instructions in the following guide.

This repo is developed based on Spatio-Temporal Segmentation and O-CNN.

Please also consider citing the corresponding papers.

@article{CSN:2023,
  author  = {Marios Loizou and Siddhant Garg and Dmitry Petrov and Melinos Averkiou and Evangelos Kalogerakis},
  title   = {{Cross-Shape Attention for Part Segmentation of 3D Point Clouds}},
  journal = {Computer Graphics Forum (Proc. SGP)},
  year    = {2023},
  volume  = {42},
  issue   = {5}
}