Saved in:
Bibliographic Details
Main Authors: Chang, Chun-Peng, Wang, Shaoxiang, Pagani, Alain, Gavrila, Dariu, Caesar, Holger
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.15923
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911688386674688
author Chang, Chun-Peng
Wang, Shaoxiang
Pagani, Alain
Gavrila, Dariu
Caesar, Holger
author_facet Chang, Chun-Peng
Wang, Shaoxiang
Pagani, Alain
Gavrila, Dariu
Caesar, Holger
contents Modern image encoders achieve high generalization by decoupling semantic meaning from resolution, an ability yet to be fully realized in the 3D domain. We investigate the failure of 3D point cloud encoders to achieve similar generalization and find that existing models are highly sensitive to sampling resolution and scale changes, leading to significant performance degradation. This sensitivity is a major bottleneck for real-world deployment in robotics, as it suggests models overfit to specific quantization densities and object scales rather than learning invariant semantic features. To mitigate this dependency, we propose Invaria, a point cloud encoder that achieves scale and density invariance through next-resolution prediction and receptive field calibration. While our objective is not the explicit generation of high-resolution point clouds, we find that this training objective encourages the model to learn robust, structural invariants. The resulting encoder achieves significant performance gains during resolution shifts while maintaining high efficiency through a compact model size and reduced token requirements. Specifically, on ScanNet, Invaria achieves a 56.0\% higher mIoU at 3$\times$ lower resolution and a 20\% improvement when the objects scale is reduced by a factor of 3. These gains are achieved with a 45\% smaller model size and an average reduction of 40\% in input tokens.
format Preprint
id arxiv_https___arxiv_org_abs_2605_15923
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Invaria: Learning Scale and Density Invariance in Point Clouds via Next-Resolution Prediction
Chang, Chun-Peng
Wang, Shaoxiang
Pagani, Alain
Gavrila, Dariu
Caesar, Holger
Computer Vision and Pattern Recognition
Modern image encoders achieve high generalization by decoupling semantic meaning from resolution, an ability yet to be fully realized in the 3D domain. We investigate the failure of 3D point cloud encoders to achieve similar generalization and find that existing models are highly sensitive to sampling resolution and scale changes, leading to significant performance degradation. This sensitivity is a major bottleneck for real-world deployment in robotics, as it suggests models overfit to specific quantization densities and object scales rather than learning invariant semantic features. To mitigate this dependency, we propose Invaria, a point cloud encoder that achieves scale and density invariance through next-resolution prediction and receptive field calibration. While our objective is not the explicit generation of high-resolution point clouds, we find that this training objective encourages the model to learn robust, structural invariants. The resulting encoder achieves significant performance gains during resolution shifts while maintaining high efficiency through a compact model size and reduced token requirements. Specifically, on ScanNet, Invaria achieves a 56.0\% higher mIoU at 3$\times$ lower resolution and a 20\% improvement when the objects scale is reduced by a factor of 3. These gains are achieved with a 45\% smaller model size and an average reduction of 40\% in input tokens.
title Invaria: Learning Scale and Density Invariance in Point Clouds via Next-Resolution Prediction
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2605.15923