LaLaLoc: latent layout localisation in dynamic, unvisited environments
We present LaLaLoc to localise in environments without the need for prior visitation, and in a manner that is robust to large changes in scene appearance, such as a full rearrangement of furniture. Specifically, LaLaLoc performs localisation through latent representations of room layout. LaLaLoc lea...
| Main Authors: | , , |
|---|---|
| Format: | Conference item |
| Sprog: | English |
| Udgivet: |
IEEE
2022
|
| _version_ | 1826259239561068544 |
|---|---|
| author | Howard-Jenkins, H Ruiz-Sarmiento, J-R Prisacariu, VA |
| author_facet | Howard-Jenkins, H Ruiz-Sarmiento, J-R Prisacariu, VA |
| author_sort | Howard-Jenkins, H |
| collection | OXFORD |
| description | We present LaLaLoc to localise in environments without the need for prior visitation, and in a manner that is robust to large changes in scene appearance, such as a full rearrangement of furniture. Specifically, LaLaLoc performs localisation through latent representations of room layout. LaLaLoc learns a rich embedding space shared between RGB panoramas and layouts inferred from a known floor plan that encodes the structural similarity between locations. Further, LaLaLoc introduces direct, cross-modal pose optimisation in its latent space. Thus, LaLaLoc enables fine-grained pose estimation in a scene without the need for prior visitation, as well as being robust to dynamics, such as a change in furniture configuration. We show that in a domestic environment LaLaLoc is able to accurately localise a single RGB panorama image to within 8.3cm, given only a floor plan as a prior. |
| first_indexed | 2024-03-06T18:46:44Z |
| format | Conference item |
| id | oxford-uuid:0ec3fefa-c8e5-40ec-8fd9-5e6f56dde5cf |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-06T18:46:44Z |
| publishDate | 2022 |
| publisher | IEEE |
| record_format | dspace |
| spelling | oxford-uuid:0ec3fefa-c8e5-40ec-8fd9-5e6f56dde5cf2022-03-26T09:47:47ZLaLaLoc: latent layout localisation in dynamic, unvisited environmentsConference itemhttp://purl.org/coar/resource_type/c_5794uuid:0ec3fefa-c8e5-40ec-8fd9-5e6f56dde5cfEnglishSymplectic ElementsIEEE2022Howard-Jenkins, HRuiz-Sarmiento, J-RPrisacariu, VAWe present LaLaLoc to localise in environments without the need for prior visitation, and in a manner that is robust to large changes in scene appearance, such as a full rearrangement of furniture. Specifically, LaLaLoc performs localisation through latent representations of room layout. LaLaLoc learns a rich embedding space shared between RGB panoramas and layouts inferred from a known floor plan that encodes the structural similarity between locations. Further, LaLaLoc introduces direct, cross-modal pose optimisation in its latent space. Thus, LaLaLoc enables fine-grained pose estimation in a scene without the need for prior visitation, as well as being robust to dynamics, such as a change in furniture configuration. We show that in a domestic environment LaLaLoc is able to accurately localise a single RGB panorama image to within 8.3cm, given only a floor plan as a prior. |
| spellingShingle | Howard-Jenkins, H Ruiz-Sarmiento, J-R Prisacariu, VA LaLaLoc: latent layout localisation in dynamic, unvisited environments |
| title | LaLaLoc: latent layout localisation in dynamic, unvisited environments |
| title_full | LaLaLoc: latent layout localisation in dynamic, unvisited environments |
| title_fullStr | LaLaLoc: latent layout localisation in dynamic, unvisited environments |
| title_full_unstemmed | LaLaLoc: latent layout localisation in dynamic, unvisited environments |
| title_short | LaLaLoc: latent layout localisation in dynamic, unvisited environments |
| title_sort | lalaloc latent layout localisation in dynamic unvisited environments |
| work_keys_str_mv | AT howardjenkinsh lalaloclatentlayoutlocalisationindynamicunvisitedenvironments AT ruizsarmientojr lalaloclatentlayoutlocalisationindynamicunvisitedenvironments AT prisacariuva lalaloclatentlayoutlocalisationindynamicunvisitedenvironments |