期刊
2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR)
卷 -, 期 -, 页码 11112-11121出版社
IEEE COMPUTER SOC
DOI: 10.1109/CVPR52688.2022.01084
关键词
-
LASER is an image-based Monte Carlo Localization (MCL) framework that utilizes latent space rendering to present 2D pose hypotheses directly into a geometrically-structured latent space. By dynamically determining viewing ray features through a codebook scheme and applying metric learning, LASER achieves state-of-the-art performance on large-scale indoor localization datasets for both panorama and perspective image queries.
We present LASER, an image-based Monte Carlo Localization (MCL) framework for 2D floor maps. LASER introduces the concept of latent space rendering, where 2D pose hypotheses on the floor map are directly rendered into a geometrically-structured latent space by aggregating viewing ray features. Through a tightly coupled rendering codebook scheme, the viewing ray features are dynamically determined at rendering-time based on their geometries (i.e. length, incident-angle), endowing our representation with view-dependent fine-grain variability. Our codebook scheme effectively disentangles feature encoding from rendering, allowing the latent space rendering to run at speeds above 10KHz. Moreover, through metric learning, our geometrically-structured latent space is common to both pose hypotheses and query images with arbitrary field of views. As a result, LASER achieves state-of-the-art performance on large-scale indoor localization datasets (i.e. ZInD [5] and Structured3D [38]) for both panorama and perspective image queries, while significantly outperforming existing learning-based methods in speed.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据