Journal article · 2026 · Open access

2D Gaussian Splatting for Semantic Segmentation via Discrete-to-Continuous Representation toward Boundary-Preserving Feature Upsampling

Gaocheng Zhang, Qi Li, Longxi Zheng, Junjie Huang, Jiaxin Cai, Xingsi Xue, Wenxi Liu, Chunhua Shen

IEEE Transactions on Pattern Analysis and Machine Intelligence, PP, 1–17 · Published 12 November 2026 · DOI 10.1109/tpami.2026.3753018

Open full textCite this paper

In brief

“2D Gaussian Splatting for Semantic Segmentation via Discrete-to-Continuous Representation toward Boundary-Preserving Feature Upsampling” is a 2026 journal article by Gaocheng Zhang and 7 co-authors, published in IEEE Transactions on Pattern Analysis and Machine Intelligence. OpenAlex counts 0 citations as of 18 November 2026. It is open access under a CC BY licence: the full text is free to read at IEEE Transactions on Pattern Analysis and Machine Intelligence.

Citations
0
Authors
8

Abstract

High-quality semantic segmentation requires not only rich semantic understanding but also precise delineation of object boundaries. While prevailing methods rely on discrete feature representations that often lead to boundary misalignment, jagged contours, and blurry edges along complex contours, we present GSSeg, a novel semantic segmentation paradigm based on Gaussian Splatting that has been successfully applied to 3D scene reconstruction. Specifically, GSSeg reformulates the discrete feature space as continuous Gaussian fields, thereby rendering spatial continuity modeling to produce high-quality segmentation results. However, directly adapting 2DGS to existing pipelines suffers from the structural granularity loss in low-resolution features and the inherent low-pass limitations of Gaussian rendering. To address the insufficient structural granularity available for discrete-to-continuous transformation, we propose the cross-layer neighborhood interaction scheme utilizing alternating global scanning, which establishes cross-scale contextual cues to enhance adaptive Gaussian kernel generation. Additionally, addressing the inherent low-pass limitations of Gaussian rendering, we introduce spatial uncertainty-aware resampling driven by resolution bias estimation, which selectively compensates for high-frequency details attenuated during Gaussian rendering. Furthermore, we introduce an effective Gaussian kernel initialization strategy that approximates Gaussian radial basis function interpolation, ensuring model training stability. Comprehensive experiments on standard benchmarks, especially high-resolution matting-level dataset MaSS13K, demonstrate that GSSeg achieves state-of-the-art performance.

Cite this paper

Formatted from the catalogue record. Check names, capitals and page numbers against the paper before you submit.

APA 7

Zhang, G., Li, Q., Zheng, L., Huang, J., Cai, J., Xue, X., Liu, W., & Shen, C. (2026). 2D Gaussian Splatting for Semantic Segmentation via Discrete-to-Continuous Representation toward Boundary-Preserving Feature Upsampling. IEEE Transactions on Pattern Analysis and Machine Intelligence, PP, 1–17. https://doi.org/10.1109/tpami.2026.3753018

MLA 9

Zhang, Gaocheng, et al. “2D Gaussian Splatting for Semantic Segmentation via Discrete-to-Continuous Representation toward Boundary-Preserving Feature Upsampling.” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. PP, 2026, pp. 1–17. https://doi.org/10.1109/tpami.2026.3753018.

Chicago

Zhang, Gaocheng, Qi Li, Longxi Zheng, Junjie Huang, Jiaxin Cai, Xingsi Xue, Wenxi Liu, and Chunhua Shen. 2026. “2D Gaussian Splatting for Semantic Segmentation via Discrete-to-Continuous Representation toward Boundary-Preserving Feature Upsampling.” IEEE Transactions on Pattern Analysis and Machine Intelligence PP: 1–17. https://doi.org/10.1109/tpami.2026.3753018.

Harvard

Zhang, G. et al. (2026) ‘2D Gaussian Splatting for Semantic Segmentation via Discrete-to-Continuous Representation toward Boundary-Preserving Feature Upsampling’, IEEE Transactions on Pattern Analysis and Machine Intelligence, PP, pp. 1–17. doi:10.1109/tpami.2026.3753018.

IEEE

G. Zhang et al., “2D Gaussian Splatting for Semantic Segmentation via Discrete-to-Continuous Representation toward Boundary-Preserving Feature Upsampling,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. PP, pp. 1–17, 2026, doi: 10.1109/tpami.2026.3753018.

BibTeX

@article{zhang2026gaussian,
  title     = {{2D Gaussian Splatting for Semantic Segmentation via Discrete-to-Continuous Representation toward Boundary-Preserving Feature Upsampling}},
  author    = {Zhang, Gaocheng and Li, Qi and Zheng, Longxi and Huang, Junjie and Cai, Jiaxin and Xue, Xingsi and Liu, Wenxi and Shen, Chunhua},
  journal   = {IEEE Transactions on Pattern Analysis and Machine Intelligence},
  year      = {2026},
  volume    = {PP},
  pages     = {1--17},
  doi       = {10.1109/tpami.2026.3753018},
  url       = {https://doi.org/10.1109/tpami.2026.3753018}
}

Download for a reference manager: BibTeX (.bib) · RIS for Zotero, Mendeley and EndNote (.ris)

Topic and keywords

Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications

  • 3D Gaussian Splatting
  • semantic segmentation
  • boundary preservation
  • feature upsampling

Work with this paper

Most-cited papers in Computer Science

All highly cited Computer Science papers

Questions about this paper

How do I cite this paper in APA style?

Zhang, G., Li, Q., Zheng, L., Huang, J., Cai, J., Xue, X., Liu, W., & Shen, C. (2026). 2D Gaussian Splatting for Semantic Segmentation via Discrete-to-Continuous Representation toward Boundary-Preserving Feature Upsampling. IEEE Transactions on Pattern Analysis and Machine Intelligence, PP, 1–17. https://doi.org/10.1109/tpami.2026.3753018

How many times has this paper been cited?

0 times, according to OpenAlex on 18 November 2026. Google Scholar and Scopus count different sets of sources, so their numbers differ.

Is this paper free to read?

Yes. It is open access at IEEE Transactions on Pattern Analysis and Machine Intelligence: sign in and open the full text from this page.

Who wrote this paper?

Gaocheng Zhang, Qi Li, Longxi Zheng, Junjie Huang, Jiaxin Cai, Xingsi Xue, Wenxi Liu, Chunhua Shen.

What is the DOI of this paper?

10.1109/tpami.2026.3753018, permanently at https://doi.org/10.1109/tpami.2026.3753018. Sign in and open it from this page.

Bibliographic data from OpenAlex (CC0), last checked 18 November 2026. ResearchGuru does not host this paper; once you are signed in, the links open it at the publisher or the repository that does.