Stream3D-VLM: Online 3D Spatial Understanding with Incremental Geometry Priors

Hugging Face Daily Papers Papers

Summary

Stream3D-VLM is an online 3D vision-language model that enables real-time spatial understanding from streaming video by incrementally integrating geometry priors and using geometry-adaptive voxel compression, outperforming existing models on 3D spatial understanding tasks.

Despite advances in 3D scene understanding, existing 3D Large Multimodal Models operate in offline settings, requiring complete scene observations or predefined video clips. In this paper, we present an online 3D vision-language model that enables real-time spatial understanding from streaming video. Our approach adopts an autoregressive streaming control modeling based on the LLM's next-token prediction objective to learn when to respond, and employs a lightweight Visual-Spatial Feature Integration (VSFI) module to incrementally inject temporally aligned geometry priors into the visual stream. To alleviate long-context decoding overhead, we propose a plug-and-play Geometry-Adaptive Voxel Compression (GAVC) module for efficient visual token compression. To address the scarcity of streaming 3D-language data, we further develop a scalable data generation pipeline that curates over 1M online spatio-temporal 3D QA pairs and establishes a comprehensive benchmark spanning 29 tasks. Extensive experiments show that our approach significantly outperforms both proprietary and open-source models across online and offline 3D spatial understanding, reasoning, and grounding tasks. The project page is available at https://stream3d-vlm.github.io/
Original Article
View Cached Full Text

Cached at: 06/08/26, 07:14 AM

Paper page - Stream3D-VLM: Online 3D Spatial Understanding with Incremental Geometry Priors

Source: https://huggingface.co/papers/2606.06891

Abstract

An online 3D vision-language model enables real-time spatial understanding from streaming video using autoregressive control modeling and efficient visual token compression.

Despite advances in 3D scene understanding, existing3D Large Multimodal Modelsoperate in offline settings, requiring complete scene observations or predefined video clips. In this paper, we present an online 3D vision-language model that enables real-time spatial understanding from streaming video. Our approach adopts anautoregressive streaming control modelingbased on the LLM’snext-token prediction objectiveto learn when to respond, and employs a lightweightVisual-Spatial Feature Integration(VSFI) module to incrementally inject temporally aligned geometry priors into the visual stream. To alleviate long-context decoding overhead, we propose a plug-and-playGeometry-Adaptive Voxel Compression(GAVC) module for efficient visual token compression. To address the scarcity of streaming 3D-language data, we further develop a scalable data generation pipeline that curates over 1M onlinespatio-temporal 3D QA pairsand establishes a comprehensive benchmark spanning 29 tasks. Extensive experiments show that our approach significantly outperforms both proprietary and open-source models across online and offline 3D spatial understanding, reasoning, and grounding tasks. The project page is available at https://stream3d-vlm.github.io/

View arXiv pageView PDFProject pageGitHub3Add to collection

Community

Stream3D-VLM is an online 3D vision-language model that supports real-time spatial understanding and interaction directly from streaming video. By incrementally integrating geometry priors and employing geometry-adaptive voxel compression, our approach enables efficient and continuous 3D scene comprehension without requiring offline processing or complete scene observations.

Upload images, audio, and videos by dragging in the text input, pasting, orclicking here.

Tap or paste here to upload images

Get this paper in your agent:

hf papers read 2606\.06891

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper1

#### JonnyYu828/Stream3D-VLM-4B Image-Text-to-Text• 5B• Updatedabout 3 hours ago • 56 • 2

Datasets citing this paper2

#### JonnyYu828/Stream3D-1M-Dataset Preview• Updatedabout 3 hours ago • 32 • 1 #### JonnyYu828/Stream3D-Bench Viewer• Updatedabout 3 hours ago • 10k • 18 • 1

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2606.06891 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

OVO-S-Bench: A Hierarchical Benchmark for Streaming Spatial Intelligence in Multimodal LLMs

Papers with Code Trending

OVO-S-Bench introduces a comprehensive human-annotated benchmark of 1,680 questions across 348 videos to evaluate streaming spatial intelligence in multimodal LLMs, revealing that even the best model (Gemini-3.1-Pro) trails human experts by 27 points. The benchmark exposes key limitations including allocentric mapping as a major bottleneck and chain-of-thought reasoning amplifying spatial errors.