Inferring object properties from human interaction and transferring them to new motions

作     者：Qian Zheng Weikai Wu Hanting Pan Niloy Mitra Daniel Cohen-Or Hui Huang Qian Zheng;Weikai Wu;Hanting Pan;Niloy Mitra;Daniel Cohen-Or;Hui Huang

作者机构：Shenzhen UniversityShenzhenChina University College LondonUK. Tel Aviv UniversityTel-AvivIsrael 

出 版 物：《Computational Visual Media》 (计算可视媒体（英文版）)

年 卷 期：2021年第7卷第3期

页      面：375-392页

核心收录：

学科分类：08[工学] 080203[工学-机械设计及理论] 0802[工学-机械工程] 

基　　金：supported in part by Shenzhen Innovation Program(JCYJ20180305125709986) National Natural Science Foundation of China(61861130365,61761146002) GD Science and Technology Program(2020A0505100064,2015A030312015) DEGP Key Project(2018KZDXM058) 

主　　题：human interaction motion object property inference motion analysis motion synthesis 

摘      要：Humans regularly interact with their surrounding *** interactions often result in strongly correlated motions between humans and the interacting *** thus ask:Is it possible to infer object properties from skeletal motion alone,even without seeing the interacting object itself?In this paper,we present a fine-grained action recognition method that learns to infer such latent object properties from human interaction motion *** inference allows us to disentangle the motion from the object property and transfer object properties to a given *** collected a large number of videos and 3 D skeletal motions of performing actors using an inertial motion capture *** analyzed similar actions and learned subtle differences between them to reveal latent properties of the interacting *** particular,we learned to identify the interacting object,by estimating its weight,or its *** results clearly demonstrate that motions and interacting objects are highly correlated and that related object latent properties can be inferred from 3 D skeleton sequences alone,leading to new synthesis possibilities for motions involving human *** dataset is available at http://***/research/2020/***.