mlsanityData to 7 Oct 2026

Dual-Branch Center-Surrounding Contrast: Rethinking Contrastive Learning for 3D Point Clouds

arXiv 2512.08673 ↗First posted 9 Dec 2025cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

Most existing self-supervised learning (SSL) approaches for 3D point clouds are dominated by generative methods based on Masked Autoencoders (MAE). However, these generative methods have been proven to struggle to capture high-level discriminative features effectively, leading to poor performance on linear probing and other downstream tasks. In contrast, contrastive methods excel in discriminative feature representation and generalization ability on image data. Despite this, contrastive learning (CL) in 3D data remains scarce. Besides, simply applying CL methods designed for 2D data to 3D fails to effectively learn 3D local details. To address these challenges, we propose a novel Dual-Branch Center-Surrounding Contrast (CSCon) framework. Specifically, we apply masking to the center and surrounding parts separately, constructing dual-branch inputs with center-biased and surrounding-biased representations to better capture rich geometric information. Meanwhile, we introduce a patch-level contrastive loss to further enhance both high-level information and local sensitivity. Under the FULL and ALL protocols, CSCon achieves performance comparable to generative methods; under the MLP-LINEAR, MLP-3, and ONLY-NEW protocols, our method attains state-of-the-art results, even surpassing cross-modal approaches. In particular, under the MLP-LINEAR protocol, our method outperforms the baseline (Point-MAE) by 7.9%, 6.7%, and 10.3% on the three variants of ScanObjectNN, respectively. The code will be made publicly available.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 52 of its 59 references that are in the map. First posted on arXiv 9 Dec 2025.

Builds on

52 of 59 references.

References · 52

  1. 2017Attention Is All You Needcited by 59,932
  2. 2020A Simple Framework for Contrastive Learning of Visual Representationscited by 11,979
  3. 2018Representation Learning with Contrastive Predictive Codingcited by 7,662
  4. 2021Masked Autoencoders Are Scalable Vision Learnerscited by 7,339
  5. 2016PointNet: Deep Learning on Point Sets for 3D Classification and Segmentationcited by 5,223
  6. 2016SGDR: Stochastic Gradient Descent with Warm Restartscited by 4,993
  7. 2017PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Spacecited by 4,384
  8. 2015ShapeNet: An Information-Rich 3D Model Repositorycited by 2,880
  9. 2021BEiT: BERT Pre-Training of Image Transformerscited by 2,360
  10. 2018Dynamic Graph CNN for Learning on Point Cloudscited by 2,220
  11. 2020Understanding Contrastive Representation Learning through Alignment and Uniformity on the Hyperspherecited by 1,393
  12. 2022DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object Detectioncited by 1,306
38 more
  1. 2021SimMIM: A Simple Framework for Masked Image Modelingcited by 1,147
  2. 2022data2vec: A General Framework for Self-supervised Learning in Speech, Vision and Languagecited by 718
  3. 2021Point-BERT: Pre-training 3D Point Cloud Transformers with Masked Point Modelingcited by 647
  4. 2021Masked Feature Prediction for Self-Supervised Visual Pre-Trainingcited by 573
  5. 2019Revisiting Point Cloud Classification: A New Benchmark Dataset and Classification Model on Real-World Datacited by 532
  6. 2020PointContrast: Unsupervised Pre-training for 3D Point Cloud Understandingcited by 502
  7. 2022Masked Autoencoders for Point Cloud Self-supervised Learningcited by 456
  8. 2022Rethinking Network Design and Local Geometry in Point Cloud: A Simple Residual MLP Frameworkcited by 376
  9. 2020Knowledge Distillation and Student-Teacher Learning for Visual Intelligence: A Review and New Outlookscited by 312
  10. 2022Context Autoencoder for Self-Supervised Representation Learningcited by 297
  11. 2022Point-M2AE: Multi-scale Masked Autoencoders for Hierarchical Point Cloud Pre-trainingcited by 263
  12. 2021PeCo: Perceptual Codebook for BERT Pre-training of Vision Transformerscited by 204
  13. 2022CrossPoint: Self-Supervised Cross-Modal Contrastive Learning for 3D Point Cloud Understandingcited by 198
  14. 2020Unsupervised Point Cloud Pre-Training via Occlusion Completioncited by 186
  15. 2021Spatio-temporal Self-Supervised Representation Learning for 3D Point Cloudscited by 180
  16. 2023Contrast with Reconstruct: Contrastive 3D Representation Learning Guided by Generative Pretrainingcited by 162
  17. 2022Masked Discrimination for Self-Supervised Learning on Point Cloudscited by 155
  18. 2021Revisiting Point Cloud Shape Classification with a Simple and Effective Baselinecited by 145
  19. 2022Learning 3D Representations from 2D Pre-trained Models via Image-to-Point Masked Autoencoderscited by 143
  20. 2020MVTN: Multi-View Transformation Network for 3D Shape Recognitioncited by 140
  21. 2022Autoencoders as Cross-Modal Teachers: Can Pretrained 2D Image Transformers Help 3D Representation Learning?cited by 121
  22. 2023PointGPT: Auto-regressively Generative Pre-training from Point Cloudscited by 115
  23. 2022Rank-N-Contrast: Learning Continuous Representations for Regressioncited by 74
  24. 2020Info3D: Representation Learning on 3D Objects using Mutual Information Maximization and Contrastive Learningcited by 68
  25. 2020Self-Supervised Few-Shot Learning on Point Cloudscited by 66
  26. 2021Self-Contrastive Learning with Hard Negative Sampling for Self-supervised Point Cloud Learningcited by 52
  27. 2023Joint-MAE: 2D-3D Joint Masked Autoencoders for 3D Point Cloud Pre-trainingcited by 52
  28. 2023Point Cloud Pre-training with Diffusion Modelscited by 47
  29. 2023Towards Compact 3D Representations via Point Feature Enhancement Masked Autoencoderscited by 42
  30. 2023Take-A-Photo: 3D-to-2D Generative Pre-training of Point Cloud Modelscited by 25
  31. 2024PCP-MAE: Learning to Predict Centers for Point Masked Autoencoderscited by 23
  32. 2024Bringing Masked Autoencoders Explicit Contrastive Properties for Point Cloud Self-Supervised Learningcited by 16
  33. 2024RI-MAE: Rotation-Invariant Masked AutoEncoders for Self-Supervised Point Cloud Representation Learningcited by 15
  34. 2023Patch-Level Contrasting without Patch Correspondence for Accurate and Dense Contrastive Representation Learningcited by 14
  35. 2025Towards More Diverse and Challenging Pre-training for Point Cloud Learning: Self-Supervised Cross Reconstruction with Decoupled Viewscited by 7
  36. 2021MaskSplit: Self-supervised Meta-learning for Few-shot Semantic Segmentationcited by 6
  37. 2025UniPre3D: Unified Pre-training of 3D Point Cloud Models with Cross-Modal Gaussian Splattingcited by 6
  38. 2025A Cross Branch Fusion-Based Contrastive Learning Framework for Point Cloud Self-supervised Learningcited by 1

50 most cited of 52 All 52 (JSON)

Built on by

Not cited yet.