Body height and weight estimation from a single non-frontal face image suffers from poor performance due to large face pose variance and lack of labeled data.In this paper,we propose a face-based body height and weigh...Body height and weight estimation from a single non-frontal face image suffers from poor performance due to large face pose variance and lack of labeled data.In this paper,we propose a face-based body height and weight estimation method that leverages auxiliary tasks and pose disentanglement to address these issues.Specifically,inspired by the relevance of gender,age,height and weight estimation tasks,we employ gender and age estimation as auxiliary tasks to improve the performance of primary tasks,i.e.,height and weight estimation.Besides,we remove the pose-relevant feature from input to further promote the performance of both primary tasks and auxiliary tasks.Extensive experiments are conducted on both small-and large-pose datasets,demonstrating the superiority of the proposed method.展开更多
基金partially supported by the National Natural Science Foundation of China(No.U2336213)Strategic Priority Research Program of the Chinese Academy of Sciences(No.XDB0680202)+2 种基金Beijing Nova Program(20230484368)Suzhou Frontier Technology Research Project(No.SYG202325)Youth Innovation Promotion Association CAS.
文摘Body height and weight estimation from a single non-frontal face image suffers from poor performance due to large face pose variance and lack of labeled data.In this paper,we propose a face-based body height and weight estimation method that leverages auxiliary tasks and pose disentanglement to address these issues.Specifically,inspired by the relevance of gender,age,height and weight estimation tasks,we employ gender and age estimation as auxiliary tasks to improve the performance of primary tasks,i.e.,height and weight estimation.Besides,we remove the pose-relevant feature from input to further promote the performance of both primary tasks and auxiliary tasks.Extensive experiments are conducted on both small-and large-pose datasets,demonstrating the superiority of the proposed method.
