ICM-3D: Instantiated Category Modeling for 3D Instance Segmentation
- 30 August 2021
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 7 (1) , 57-64
Abstract
Separating 3D point clouds into individual instances is an important task for 3D vision. It is challenging due to the unknown and varying number of instances in a scene. Existing deep learning based works focus on a two-step pipeline: first learn a feature embedding and then cluster the points. Such a two-step pipeline leads to disconnected intermediate objectives. In this paper, we propose an integrated reformulation of 3D instance segmentation as a per-point classification problem. We propose ICM-3D, a single-step method to segment 3D instances via instantiated categorization. The augmented category information is automatically constructed from 3D spatial positions. We conduct extensive experiments to verify the effectiveness of ICM-3D and show that it obtains inspiring performance across multiple frameworks, backbones and benchmarks.Keywords
All Related Versions
This publication has 28 references indexed in Scilit:
- SGPN: Similarity Group Proposal Network for 3D Point Cloud Instance Segmentation2018
- 3D Semantic Segmentation with Submanifold Sparse Convolutional Networks2018
- SceneCut: Joint Geometric and Object Segmentation for Indoor Scenes2018
- Mask R-CNN2017
- Fully Convolutional Instance-Aware Semantic Segmentation2017
- ScanNet: Richly-Annotated 3D Reconstructions of Indoor Scenes2017
- V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation2016
- 3D Semantic Parsing of Large-Scale Indoor Spaces2016
- Incremental attention-driven object segmentation2014
- Mean shift: a robust approach toward feature space analysisIEEE Transactions on Pattern Analysis and Machine Intelligence, 2002