arXiv:2605.30111v1 Announce Type: new 
Abstract: Point cloud segmentation is a fundamental task in 3D scene understanding. Its progress is constrained by the high cost and time required for dense 3D annotations, making labeled samples difficult to obtain. Beyond annotation scarcity, different sensing modalities face inherent limitations. 2D images provide rich texture and appearance cues, yet they lack explicit depth and geometric structure. In contrast, 3D point clouds capture accurate spatial geometry but are sparse and contain no texture information. As a result, relying on a single modality restricts the richness of learned representations and weakens generalization. Although recent multi-modal methods that combine 3D point clouds with 2D images have demonstrated strong performance in tasks such as classification and retrieval, they typically depend on large-scale labeled datasets and have not been fully exploited for data-efficient dense prediction. To address these limitations, we propose a novel cross-modal knowledge distillation framework, xModel-KD, for 3D point cloud segmentation. Our method exploits the complementary strengths of 2D texture and 3D geometry by learning unified per-point representations through cross-modal alignment. Specifically, we design a cross-modal fusion encoder trained with a contrastive objective that enforces feature consistency between corresponding 2D and 3D representations across multiple views. By integrating powerful pre-trained backbones with a targeted fusion strategy, the proposed framework effectively transfers appearance cues from images to geometry-aware point features. Experimental results show that cross-modal fusion achieves a 2% absolute improvement in mIoU over a LiDAR-only baseline, demonstrating the benefit of leveraging complementary multi-modal information for scalable and annotation-efficient 3D scene understanding.

تقدم دراسة جديدة xModel-KD، وهو إطار عمل لتقطير المعرفة عبر الأنماط مصمم لتحسين إدراك المشاهد ثلاثية الأبعاد باستخدام بيانات LiDAR. تتناول هذه الطريقة التحديات المتعلقة بتجزئة سحب النقاط، والتي غالبًا ما تكون محدودة بسبب ندرة العينات المعلّمة والقيود الكامنة في أنماط الاستشعار المختلفة.

Un nuevo estudio presenta xModel-KD, un marco de destilación de conocimiento cross-modal diseñado para mejorar la percepción de escenas en 3D utilizando datos de LiDAR. Este enfoque aborda los desafíos de la segmentación de nubes de puntos, que a menudo se ve limitada por la escasez de muestras etiquetadas y las limitaciones inherentes de las diferentes modalidades de detección.

Une nouvelle étude présente xModel-KD, un cadre de distillation de connaissances cross-modal conçu pour améliorer la perception des scènes en 3D à l'aide de données LiDAR. Cette approche répond aux défis de la segmentation des nuages de points, souvent limitée par la rareté des échantillons étiquetés et les limitations inhérentes des différentes modalités de détection.

A new study introduces xModel-KD, a cross-modal knowledge distillation framework designed to enhance 3D scene perception using LiDAR data. This approach addresses the challenges of point cloud segmentation, which is often limited by the scarcity of labeled samples and the inherent limitations of different sensing modalities.

xModel-KD: Cross-modal Knowledge Distillation for 3D Scene Perception using LiDAR

arXiv:2603.05965v3 Announce Type: replace-cross 
Abstract: We present PROBE (PRobabilistic Occupancy BEV Encoding), a learning-free LiDAR place recognition descriptor that models each BEV cell's occupancy as a Bernoulli random variable. Rather than relying on discrete point-cloud perturbations, PROBE analytically marginalizes over continuous Cartesian translations via the polar Jacobian, yielding a distance-adaptive angular uncertainty $\sigma_\theta = \sigma_t / r$ in $\mathcal{O}(R{\cdot}S)$ time. The primary parameter $\sigma_t$ represents the expected translational uncertainty in meters, a sensor-independent physical quantity that enhances cross-sensor generalization while reducing the need for extensive per-dataset tuning. Pairwise similarity combines a Bernoulli-KL Jaccard with exponential uncertainty gating and FFT-based height cosine similarity for rotation alignment. Evaluated on four datasets spanning four diverse LiDAR types, PROBE achieves the highest accuracy among handcrafted descriptors in multi-session evaluation and competitive single-session performance relative to both handcrafted and supervised baselines. The source code and supplementary materials are available at https://sites.google.com/view/probe-pr.

قدم الباحثون نظام PROBE (ترميز احتمالي للاحتلال من منظور الطائر) كأداة للتعرف على الأماكن باستخدام LiDAR، حيث يقوم بنمذجة الاحتلال في خلايا العرض العلوي (BEV) كمتغيرات عشوائية من نوع برنولي، مما يعزز عدم اليقين الزاوي القابل للتكيف مع المسافة والتعميم بين المستشعرات.

Investigadores han presentado PROBE (PRobabilistic Occupancy BEV Encoding), un descriptor de reconocimiento de lugares LiDAR sin aprendizaje que modela la ocupación en celdas de vista cenital (BEV) como variables aleatorias de Bernoulli, mejorando la incertidumbre angular adaptativa a la distancia y la generalización entre sensores.

Des chercheurs ont présenté PROBE (PRobabilistic Occupancy BEV Encoding), un descripteur de reconnaissance de lieu LiDAR sans apprentissage qui modélise l'occupation dans les cellules de vue de dessus (BEV) comme des variables aléatoires de Bernoulli, améliorant l'incertitude angulaire adaptative à la distance et la généralisation inter-capteurs.

Researchers have introduced PROBE (PRobabilistic Occupancy BEV Encoding), a learning-free LiDAR place recognition descriptor that models occupancy in bird's-eye view (BEV) cells as Bernoulli random variables, enhancing distance-adaptive angular uncertainty and cross-sensor generalization.

xModel-KD: Cross-modal Knowledge Distillation for 3D Scene Perception using LiDAR

Was this article worth reading? Share it

Ready to build your own newsroom?