DiffusionPointLabel: Annotated Point Cloud Generation with Diffusion Model

Li, Tingting; Fu, Yunfei; Han, Xiaoguang; Liang, Hui; Zhang, Jian Jun; Chang, Jian

DiffusionPointLabel: Annotated Point Cloud Generation with Diffusion Model

Files

v41i7pp131-139.pdf (2.19 MB)

paper1150_supplementary_material.pdf (716.17 KB)

Date

2022

Authors

Li, Tingting
Fu, Yunfei
Han, Xiaoguang
Liang, Hui
Zhang, Jian Jun
Chang, Jian

Publisher

The Eurographics Association and John Wiley & Sons Ltd.

Abstract

Point cloud generation aims to synthesize point clouds that do not exist in supervised dataset. Generating a point cloud with certain semantic labels remains an under-explored problem. This paper proposes a formulation called DiffusionPointLabel, which completes point-label pair generation based on a DDPM generative model (Denoising Diffusion Probabilistic Model). Specifically, we use a point cloud diffusion generative model and aggregate the intermediate features of the generator. On top of this, we propose Feature Interpreter that transforms intermediate features into semantic labels. Furthermore, we employ an uncertainty measure to filter unqualified point-label pairs for a better quality of generated point cloud dataset. Coupling these two designs enables us to automatically generate annotated point clouds, especially when supervised point-labels pairs are scarce. Our method extends the application of point cloud generation models and surpasses state-of-the-art models.

CCS Concepts: Methods and Applications → Point-Based Methods

        @article{10.1111:cgf.14663
,
journal = {Computer Graphics Forum},
title = {{DiffusionPointLabel: Annotated Point Cloud Generation with Diffusion Model
}},
author = {Li, Tingting and 
Fu, Yunfei and 
Han, Xiaoguang and 
Liang, Hui and 
Zhang, Jian Jun and 
Chang, Jian
},
year = {2022
},
publisher = {The Eurographics Association and John Wiley & Sons Ltd.
},
ISSN = {1467-8659
},
DOI = {10.1111/cgf.14663
}
}