{"id":753430,"date":"2021-04-20T16:56:20","date_gmt":"2021-04-20T23:56:20","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?post_type=msr-research-item&p=753430"},"modified":"2021-06-16T08:01:56","modified_gmt":"2021-06-16T15:01:56","slug":"volumetric-mapping-for-long-term-robot-interaction-jrc-workshop-2021","status":"publish","type":"msr-video","link":"https:\/\/www.microsoft.com\/en-us\/research\/video\/volumetric-mapping-for-long-term-robot-interaction-jrc-workshop-2021\/","title":{"rendered":"Volumetric Mapping for Long-term Robot Interaction | JRC Workshop 2021"},"content":{"rendered":"
Computer Vision | Day 1
\n20 April 2021<\/p>\n
Speaker: Lukas Schmid, ETH Zurich
\n(collaboration with Cesar Cadena, Roland Siegwart, ETH Zurich and Johannes Sch\u00f6nberger, Juan Nieto, Marc Pollefeys, Microsoft)<\/p>\n
Having a high-quality map, i.e. a geometric and semantic understanding of one’s surroundings over longer periods of time, is a prerequisite for a large variety of interactive tasks in robotics and AR, such as navigation, object search, manipulation, immersive AR experiences, or service robotics. However, state-of-the-art approaches are not yet well equipped to capture long-term dynamics in changing scenes. In this talk, we present current work a novel map representation leveraging semantic scene understanding to capture temporal changes in long-term dynamic environments and to provide the agent with object-centric multi-resolution maps.<\/p>\n