德国马克斯普朗克信息研究所孙倩茹博士学术报告

专家简介:孙倩茹博士于20161月获北京大学信息科学技术学院博士学位。现任德国马克思普朗克信息研究所(Max Planck Institute for Informatics) 博士后。孙博士在国际顶级期刊和会议如PR,CVPR, NIPS等发表学术论文18篇,并担任各项国际期刊会议评审,如IJCV,TCSVT, TIP, ICCV, CVPR. 孙博士目前主要致力于统计模型在计算机视觉的应用,如基于对抗生成网络的人体图像生成以及其在社交场合中的建模和分析。孙博士在博士期间荣获北京大学中营奖学金,在博士后期间获得德国MPILise Meitner Award. 

报告题目:Person Image Generation and Applications

报告摘要:Generating realistic-looking images isof great value for many applications such as face editing, movie making anddata augmentation based on synthetic images. We proposed the novel Pose GuidedPerson Generation Network (PG2) that allows to synthesize person images inarbitrary poses. We achieved impressive results on both 128×64re-identification images and 256×256 fashion photos show that our modelgenerates high-quality images with convincing details. We proved that thesesynthetic images as augmented training data do improve the personre-identification tasks. Other applications have been made in landmark basedperson head generation from which the synthetic heads are used for personidentity obfuscation.

报告时间:2018年3月26日(周一)14:00

报告地点:山东大学青岛校区 N3-332