.png){kind=icon}
Return
PointDiffuse: A Dual-Conditional Diffusion Model for Enhanced Point Cloud Semantic Segmentation
.png){kind=icon}
PRE
AI
DOI：10.1109/tase.2026.3722737.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Diffusion probabilistic models are traditionally used to generate colors at fixed pixel positions in 2D images. Building on this, we extend diffusion models to point cloud semantic segmentation, where point positions also remain fixed, and the diffusion model generates point labels instead of colors. To accelerate the denoising process in reverse diffusion, we introduce a noisy label embedding mechanism. This approach integrates semantic information into the noisy label, providing an initial semantic reference that improves the reverse diffusion efficiency. Additionally, we propose a point frequency transformer that enhances the adjustment of high-level context in point clouds. To reduce computational complexity, we introduce the position condition into MLP and propose denoising PointNet to process the high-resolution point cloud without sacrificing geometric details. Finally, we integrate the proposed noisy label embedding, point frequency transformer and denoising PointNet in our proposed dual conditional diffusion model-based network (PointDiffuse) to perform large-scale point cloud semantic segmentation. Extensive experiments on five benchmarks demonstrate the superiority of PointDiffuse, achieving the state-of-the-art mIoU of 74.2% on S3DIS Area 5, 81.2% on S3DIS 6-fold and 64.8% on SWAN dataset. Note to Practitioners—Point cloud semantic segmentation is vital for automation in robotics, autonomous driving, and 3D inspection. Real-world point clouds are often noisy and geometrically distorted, challenging conventional segmentation methods. The proposed PointDiffuse framework tackles this by using a diffusion model to directly generate point labels, turning the segmentation task into a robust denoising process. Its Noisy Label Embedding and Point Frequency Transformer are specifically designed to handle the uncertainty and noise inherent in sensor data, improving reliability in practical applications. Current work focuses on single-dataset training; future efforts will explore multi-dataset joint training to enhance generalization across diverse environments and sensor types.
Keywords:
3D point clouds
point frequency transformer
diffusion model
Journal
IF:
6.4
Papers:
5.1K
Citations:
1.6W
Organization
View more organizations
Cited Papers
No cited papers available