TY - GEN
T1 - LASER
T2 - 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2022
AU - Min, Zhixiang
AU - Khosravan, Naji
AU - Bessinger, Zachary
AU - Narayana, Manjunath
AU - Kang, Sing Bing
AU - Dunn, Enrique
AU - Boyadzhiev, Ivaylo
N1 - Publisher Copyright:
© 2022 IEEE.
PY - 2022
Y1 - 2022
N2 - 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.
AB - 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.
KW - Pose estimation and tracking
KW - Robot vision
UR - http://www.scopus.com/inward/record.url?scp=85138965487&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85138965487&partnerID=8YFLogxK
U2 - 10.1109/CVPR52688.2022.01084
DO - 10.1109/CVPR52688.2022.01084
M3 - Conference contribution
AN - SCOPUS:85138965487
T3 - Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition
SP - 11112
EP - 11121
BT - Proceedings - 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2022
Y2 - 19 June 2022 through 24 June 2022
ER -