Cross-Attentive Multiview Fusion of Vision-Language Embeddings

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Martins, Tomas Berriel, Oswald, Martin R., Civera, Javier
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918445959872512
author Martins, Tomas Berriel
Oswald, Martin R.
Civera, Javier
author_facet Martins, Tomas Berriel
Oswald, Martin R.
Civera, Javier
contents Vision-language models have been key to the development of open-vocabulary 2D semantic segmentation. Lifting these models from 2D images to 3D scenes, however, remains a challenging problem. Existing approaches typically back-project and average 2D descriptors across views, or heuristically select a single representative one, often resulting in suboptimal 3D representations. In this work, we introduce a novel multiview transformer architecture that cross-attends across vision-language descriptors from multiple viewpoints and fuses them into a unified per-3D-instance embedding. As a second contribution, we leverage multiview consistency as a self-supervision signal for this fusion, which significantly improves performance when added to a standard supervised target-class loss. Our Cross-Attentive Multiview Fusion, which we denote with its acronym CAMFusion, not only consistently outperforms naive averaging or single-view descriptor selection, but also achieves state-of-the-art results on 3D semantic and instance classification benchmarks, including zero-shot evaluations on out-of-domain datasets.
format Preprint
id arxiv_https___arxiv_org_abs_2604_12551
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Cross-Attentive Multiview Fusion of Vision-Language Embeddings
Martins, Tomas Berriel
Oswald, Martin R.
Civera, Javier
Computer Vision and Pattern Recognition
Vision-language models have been key to the development of open-vocabulary 2D semantic segmentation. Lifting these models from 2D images to 3D scenes, however, remains a challenging problem. Existing approaches typically back-project and average 2D descriptors across views, or heuristically select a single representative one, often resulting in suboptimal 3D representations. In this work, we introduce a novel multiview transformer architecture that cross-attends across vision-language descriptors from multiple viewpoints and fuses them into a unified per-3D-instance embedding. As a second contribution, we leverage multiview consistency as a self-supervision signal for this fusion, which significantly improves performance when added to a standard supervised target-class loss. Our Cross-Attentive Multiview Fusion, which we denote with its acronym CAMFusion, not only consistently outperforms naive averaging or single-view descriptor selection, but also achieves state-of-the-art results on 3D semantic and instance classification benchmarks, including zero-shot evaluations on out-of-domain datasets.
title Cross-Attentive Multiview Fusion of Vision-Language Embeddings
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2604.12551