Newest papers

Learning from Automatic 3D Gaussian Splatting Edits for Online Bayesian Scene Change Detection

Changrui Zhu, Simon Julier · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.28490 · cited by 0 works (OpenAlex) · open-access PDF

Spanish

La detección de cambios de escena en línea requiere la integración de observaciones entrantes para determinar qué regiones han cambiado en relación con una escena de referencia. Actualizamos las creencias de cambio a medida que llegan nuevas imágenes utilizando la fusión de evidencias bayesianas generalizadas. Sin embargo, la probabilidad previa de cambios y modelos de evidencia son desconocidos previamente, y la inferencia bayesiana numérica puede ser costosa. Nos enfrentamos a estos desafíos al aprender offline de las ediciones automáticas a un modelo de 3D Gaussian Splatting (3DGS) de referencia. Estas ediciones generan vistas de entrenamiento etiquetadas sin anotaciones manuales. Junto con las observaciones de referencia inalteradas, proporcionan supervisión para los modelos de evidencia específicos de la escena del aprendizaje y la estimación de las estadísticas de píxeles efectivas

Machine translation (Meta M2M100); the English original follows.

Abstract

Online scene change detection requires integrating incoming observations to determine which regions have changed relative to a reference scene. We update change beliefs as new images arrive using generalised Bayesian evidence fusion. However, the prior probability of change and evidence models are unknown beforehand, and numerical Bayesian inference can be costly. We address these challenges by learning offline from automatic edits to a reference 3D Gaussian Splatting (3DGS) model. These edits generate labelled training views without manual annotations. Together with unchanged reference observations, they provide supervision for learning scene-specific evidence models and estimating the effective pixel prior and response statistics. Online prediction combines current evidence with historical support, while closed-form local Bayes factors, parameterized by the learned response statistics, update an accumulated 3D Gaussian field. Object-guided refinement produces the final change masks. Our method achieves state-of-the-art accuracy on PASLCD and CCS3D, with $67.64\%$ and $75.82\%$ mIoU, respectively. On PASLCD, the complete pipeline runs at $17.2$\,FPS over complete query sequences, the fastest among the evaluated online methods. Project page: https://changruizhu96.github.io/work/lego-scd/

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Change detection, Artificial intelligence, Bayesian probability, Inference, Gaussian process

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange