首页|Patent Issued for Occlusion-aware prediction of human behavior (USPTO 12094252)
Patent Issued for Occlusion-aware prediction of human behavior (USPTO 12094252)
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
Humanising Autonomy Limited (London, U nited Kingdom) has been issued patent number 12094252, according to news reporti ng originating out of Alexandria, Virginia, by NewsRx editors. The patent's inventors are Bose, Raunaq (London, GB), El Youssoufi, Wassim (Lond on, GB), Everard, James Over (Manningtree, GB), Nooteboom, Leslie Cees (London, GB), Noy, Dominic (London, GB), Pindeus, Maya Audrey Lara (London, GB), Pradeep, Yazhini Chitra (London, GB). This patent was filed on December 13, 2021 and was published online on September 17, 2024. From the background information supplied by the inventors, news correspondents o btained the following quote: "When making predictions about human behavior with computer vision-based technology, a frequent issue is that a human being in vide o frames may be occluded, as humans are often temporarily hidden from cameras be hind other objects or other people in the environment. Occlusion poses challenge s in understanding what behavior a person is likely to exhibit in the near futur e, as key points used to drive machine learning analysis of behavior are not vis ible in images input into models. Yet further, where a person is fully occluded, challenges are posed in determining whether someone is the same person when the y re-appear from the occlusion, as tracking may rely on the occluded key points. When using existing systems, occlusions often cause improper assumptions, which lead to improper predictions. For example, if a person is behind a wall and the ir legs are occluded, existing systems might assume that the person is not movin g while they actually are or assume that the person is a different person when t he occlusion is removed."
NETL
NSTL
