VL-KnG: Persistent Spatiotemporal Knowledge Graphs from Egocentric Video for Embodied Scene Understanding

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Mdfaa, Mohamad Al, Lukina, Svetlana, Akhtyamov, Timur, Nigmatzyanov, Arthur, Nalberskii, Dmitrii, Zagoruyko, Sergey, Ferrer, Gonzalo
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866908907728797696
author Mdfaa, Mohamad Al
Lukina, Svetlana
Akhtyamov, Timur
Nigmatzyanov, Arthur
Nalberskii, Dmitrii
Zagoruyko, Sergey
Ferrer, Gonzalo
author_facet Mdfaa, Mohamad Al
Lukina, Svetlana
Akhtyamov, Timur
Nigmatzyanov, Arthur
Nalberskii, Dmitrii
Zagoruyko, Sergey
Ferrer, Gonzalo
contents Vision-language models (VLMs) demonstrate strong image-level scene understanding but often lack persistent memory, explicit spatial representations, and computational efficiency when reasoning over long video sequences. We present VL-KnG, a training-free framework that constructs spatiotemporal knowledge graphs from monocular video, bridging fine-grained scene graphs and global topological graphs without 3D reconstruction. VL-KnG processes video in chunks, maintains persistent object identity via LLM-based Spatiotemporal Object Association (STOA), and answers queries via Graph-Enhanced Retrieval (GER), a hybrid of GraphRAG subgraph retrieval and SigLIP2 visual grounding. Once built, the knowledge graph eliminates the need to re-process video at query time, enabling constant-time inference regardless of video length. Evaluation across three benchmarks, OpenEQA, NaVQA, and WalkieKnowledge (our newly introduced benchmark), shows that VL-KnG matches or surpasses frontier VLMs on embodied scene understanding tasks at significantly lower query latency, with explainable, graph-grounded reasoning. Real-world robot deployment confirms practical applicability with constant-time scaling.
format Preprint
id arxiv_https___arxiv_org_abs_2510_01483
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle VL-KnG: Persistent Spatiotemporal Knowledge Graphs from Egocentric Video for Embodied Scene Understanding
Mdfaa, Mohamad Al
Lukina, Svetlana
Akhtyamov, Timur
Nigmatzyanov, Arthur
Nalberskii, Dmitrii
Zagoruyko, Sergey
Ferrer, Gonzalo
Robotics
Artificial Intelligence
Vision-language models (VLMs) demonstrate strong image-level scene understanding but often lack persistent memory, explicit spatial representations, and computational efficiency when reasoning over long video sequences. We present VL-KnG, a training-free framework that constructs spatiotemporal knowledge graphs from monocular video, bridging fine-grained scene graphs and global topological graphs without 3D reconstruction. VL-KnG processes video in chunks, maintains persistent object identity via LLM-based Spatiotemporal Object Association (STOA), and answers queries via Graph-Enhanced Retrieval (GER), a hybrid of GraphRAG subgraph retrieval and SigLIP2 visual grounding. Once built, the knowledge graph eliminates the need to re-process video at query time, enabling constant-time inference regardless of video length. Evaluation across three benchmarks, OpenEQA, NaVQA, and WalkieKnowledge (our newly introduced benchmark), shows that VL-KnG matches or surpasses frontier VLMs on embodied scene understanding tasks at significantly lower query latency, with explainable, graph-grounded reasoning. Real-world robot deployment confirms practical applicability with constant-time scaling.
title VL-KnG: Persistent Spatiotemporal Knowledge Graphs from Egocentric Video for Embodied Scene Understanding
topic Robotics
Artificial Intelligence
url https://arxiv.org/abs/2510.01483