arXiv:2512.08330v1 Announce Type: new 
Abstract: Self-supervised representation learning has shown significant improvement in Natural Language Processing and 2D Computer Vision. However, existing methods face difficulties in representing 3D data because of its unordered and uneven density. Through an in-depth analysis of mainstream contrastive and generative approaches, we find that contrastive models tend to suffer from overfitting, while 3D Mask Autoencoders struggle to handle unordered point clouds. This motivates us to learn 3D representations by sharing the merits of diffusion and contrast models, which is non-trivial due to the pattern difference between the two paradigms. In this paper, we propose \textit{PointDico}, a novel model that seamlessly integrates these methods. \textit{PointDico} learns from both denoising generative modeling and cross-modal contrastive learning through knowledge distillation, where the diffusion model serves as a guide for the contrastive model. We introduce a hierarchical pyramid conditional generator for multi-scale geometric feature extraction and employ a dual-channel design to effectively integrate local and global contextual information. \textit{PointDico} achieves a new state-of-the-art in 3D representation learning, \textit{e.g.}, \textbf{94.32\%} accuracy on ScanObjectNN, \textbf{86.5\%} Inst. mIoU on ShapeNetPart.

تم تقديم PointDico كنموذج جديد لتعلم تمثيلات ثلاثية الأبعاد المتباينة، مستفيدًا من مزايا نماذج الانتشار لتحسين عملية التعلم. تتناول هذه الطريقة التحديات التي تواجهها الأساليب الحالية في تمثيل البيانات ثلاثية الأبعاد غير المرتبة وغير المتساوية الكثافة، خاصة في سياق التعلم الذاتي الإشراف.

Se ha presentado PointDico como un nuevo modelo para el aprendizaje de representaciones 3D contrastivas, aprovechando las fortalezas de los modelos de difusión para mejorar el proceso de aprendizaje. Este enfoque aborda los desafíos que enfrentan los métodos existentes en la representación de datos 3D desordenados y de densidad desigual, especialmente en el contexto del aprendizaje auto-supervisado.

PointDico a été introduit comme un nouveau modèle pour l'apprentissage de représentations 3D contrastives, tirant parti des forces des modèles de diffusion pour améliorer le processus d'apprentissage. Cette approche répond aux défis rencontrés par les méthodes existantes dans la représentation de données 3D désordonnées et inégalement réparties, en particulier dans le contexte de l'apprentissage auto-supervisé.

PointDico has been introduced as a novel model for contrastive 3D representation learning, leveraging the strengths of diffusion models to enhance the learning process. This approach addresses the challenges faced by existing methods in representing unordered and unevenly distributed 3D data, particularly in the context of self-supervised learning.

PointDico: Contrastive 3D Representation Learning Guided by Diffusion Models

arXiv:2512.08673v1 Announce Type: new 
Abstract: 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 \textbf{C}enter-\textbf{S}urrounding \textbf{Con}trast (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 \textbf{7.9\%}, \textbf{6.7\%}, and \textbf{10.3\%} on the three variants of ScanObjectNN, respectively. The code will be made publicly available.

تم اقتراح إطار عمل جديد يسمى Dual-Branch Center-Surrounding Contrast (CSCon) لتحسين التعلم التبايني لنقاط السحاب ثلاثية الأبعاد، حيث يتناول القيود المفروضة على أساليب التعلم الذاتي التي تعتمد بشكل كبير على الأساليب التوليدية مثل أجهزة التشفير المقنعة. يهدف هذا الإطار الجديد إلى التقاط الميزات التمييزية عالية المستوى بشكل أفضل من خلال تطبيق القناع على كل من الأجزاء المركزية والمحيطة من البيانات.

Se ha propuesto un nuevo marco llamado Dual-Branch Center-Surrounding Contrast (CSCon) para mejorar el aprendizaje contrastivo de nubes de puntos 3D, abordando las limitaciones de los métodos de aprendizaje auto-supervisado que dependen en gran medida de enfoques generativos como los autoencoders enmascarados. Este nuevo marco busca capturar mejor las características discriminativas de alto nivel aplicando enmascaramiento tanto en las partes centrales como en las circundantes de los datos.

Un nouveau cadre appelé Dual-Branch Center-Surrounding Contrast (CSCon) a été proposé pour améliorer l'apprentissage contrastif des nuages de points 3D, en s'attaquant aux limitations des méthodes d'apprentissage auto-supervisé qui reposent fortement sur des approches génératives comme les autoencodeurs masqués. Ce nouveau cadre vise à mieux capturer les caractéristiques discriminatives de haut niveau en appliquant un masquage à la fois sur les parties centrales et environnantes des données.

A novel Dual-Branch Center-Surrounding Contrast (CSCon) framework has been proposed to enhance contrastive learning for 3D point clouds, addressing the limitations of existing self-supervised learning methods that rely heavily on generative approaches like Masked Autoencoders. This new framework aims to better capture high-level discriminative features by applying masking to both center and surrounding parts of the data separately.

PointDico: Contrastive 3D Representation Learning Guided by Diffusion Models

Was this article worth reading? Share it

LucidQuery AI

Deptho.ai

Artefacts.ai