Salvato in:
Dettagli Bibliografici
Autori principali: Szilagyi, Laszlo, Engelmann, Francis, Bohg, Jeannette
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:https://arxiv.org/abs/2505.08124
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866913994777821184
author Szilagyi, Laszlo
Engelmann, Francis
Bohg, Jeannette
author_facet Szilagyi, Laszlo
Engelmann, Francis
Bohg, Jeannette
contents Language-augmented scene representations hold great promise for large-scale robotics applications such as search-and-rescue, smart cities, and mining. Many of these scenarios are time-sensitive, requiring rapid scene encoding while also being data-intensive, necessitating scalable solutions. Deploying these representations on robots with limited computational resources further adds to the challenge. To address this, we introduce SLAG, a multi-GPU framework for language-augmented Gaussian splatting that enhances the speed and scalability of embedding large scenes. Our method integrates 2D visual-language model features into 3D scenes using SAM and CLIP. Unlike prior approaches, SLAG eliminates the need for a loss function to compute per-Gaussian language embeddings. Instead, it derives embeddings from 3D Gaussian scene parameters via a normalized weighted average, enabling highly parallelized scene encoding. Additionally, we introduce a vector database for efficient embedding storage and retrieval. Our experiments show that SLAG achieves an 18 times speedup in embedding computation on a 16-GPU setup compared to OpenGaussian, while preserving embedding quality on the ScanNet and LERF datasets. For more details, visit our project website: https://slag-project.github.io/.
format Preprint
id arxiv_https___arxiv_org_abs_2505_08124
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SLAG: Scalable Language-Augmented Gaussian Splatting
Szilagyi, Laszlo
Engelmann, Francis
Bohg, Jeannette
Computer Vision and Pattern Recognition
Artificial Intelligence
Robotics
Language-augmented scene representations hold great promise for large-scale robotics applications such as search-and-rescue, smart cities, and mining. Many of these scenarios are time-sensitive, requiring rapid scene encoding while also being data-intensive, necessitating scalable solutions. Deploying these representations on robots with limited computational resources further adds to the challenge. To address this, we introduce SLAG, a multi-GPU framework for language-augmented Gaussian splatting that enhances the speed and scalability of embedding large scenes. Our method integrates 2D visual-language model features into 3D scenes using SAM and CLIP. Unlike prior approaches, SLAG eliminates the need for a loss function to compute per-Gaussian language embeddings. Instead, it derives embeddings from 3D Gaussian scene parameters via a normalized weighted average, enabling highly parallelized scene encoding. Additionally, we introduce a vector database for efficient embedding storage and retrieval. Our experiments show that SLAG achieves an 18 times speedup in embedding computation on a 16-GPU setup compared to OpenGaussian, while preserving embedding quality on the ScanNet and LERF datasets. For more details, visit our project website: https://slag-project.github.io/.
title SLAG: Scalable Language-Augmented Gaussian Splatting
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Robotics
url https://arxiv.org/abs/2505.08124