Abstract
This disclosure describes techniques for device localization based on semantic information available in the environment. A semantic data source is identified by performing landmark identification using an artificial intelligence (AI) model trained to identify landmarks. Image(s) of the semantic data source are provided to a text recognition module. The detected text is provided to a semantic aware localizer. An estimated surface normal and estimated depth of the semantic data source are additionally provided as inputs to the semantic aware localizer. When device localization is being performed, live semantic information captured by the device can be utilized in conjunction with a 2D map of the environment to accurately estimate the location and spatial orientation of the device based on a comparison of the obtained semantic information with the 2D map. The localization result can be utilized in a navigation application and other applications.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Huang, Kan; Guo, Chao; and Fan, Fengtao, "Semantic Information Based Device Localization Using Artificial Intelligence", Technical Disclosure Commons, (January 20, 2025)
https://www.tdcommons.org/dpubs_series/7746