Beyond Face Recognition: Deep Face Analysis from a Micro Perspective

來(lái)源:科學(xué)技術(shù)處、電子與通信工程系發(fā)布時(shí)間:2020-07-24

【講座題目】Beyond Face Recognition: Deep Face Analysis from a Micro Perspective

【講座時(shí)間】2020年7月28日(星期二)上午10:20

【講座地點(diǎn)】網(wǎng)絡(luò)直播:https://live.bilibili.com/22339632

【主 人】韓琥,中國(guó)科學(xué)院計(jì)算技術(shù)研究所副研究員

【主講人簡(jiǎn)介】

中科院計(jì)算所副研究員,2011 年博士畢業(yè)于中科院計(jì)算所,之后分 別在美國(guó)密歇根州立大學(xué)和美國(guó)谷歌總部從事生物特征識(shí)別研究工 作,曾擔(dān)任谷歌 Abacus 項(xiàng)目核心研發(fā)成員。2015 年回到中科院計(jì)算 所工作,主要研究方向?yàn)橛?jì)算機(jī)視覺(jué)與模式識(shí)別、生物特征識(shí)別、 醫(yī)療影像分析。擔(dān)任國(guó)際期刊 Pattern Recognition 編委(AE),國(guó) 際會(huì)議 ICPR2020 領(lǐng) 域 主 席 (AC) ,以及 CVPR2020/FG2020/WACV2020/FG2019/PRCV2019 等主會(huì)特別會(huì)議/專(zhuān) 題論壇/競(jìng)賽的共同組織者。在 IEEE TPAMI/TIP/TIFS/TBIOM、PR、 CVPR、ECCV、NeurIPS、MICCAI 等領(lǐng)域權(quán)威國(guó)際期刊與會(huì)議上發(fā)表學(xué) 術(shù)論文 60 余篇,谷歌學(xué)術(shù)引用 2800 余次。作為負(fù)責(zé)人承擔(dān)國(guó)家重 點(diǎn)研發(fā)子課題、基金重點(diǎn)子課題、基金面上、中科院對(duì)外合作及企 業(yè)合作等課題 10 余項(xiàng)。研究工作 3 次獲會(huì)議最佳學(xué)生/海報(bào)論文獎(jiǎng), 3 次獲國(guó)際競(jìng)賽冠亞軍。

【內(nèi)容簡(jiǎn)介】

Benefit from the advances of deep learning methods and computing capacity, as well as the availability of large labeled dataset, face recognition performance is getting saturated, and has been widely used in a number of typical scenarios. Deep analysis of face is required to meet the requirement of new emerging application scenarios like health monitoring, ADS, and CAD of brain and mental disorders. One of the key challenges is the analysis of micro face signal in both spatial and temporal dimensions. This talk will review related progress in this area and introduce of recent work in deep face analysis with applications in facial AU recognition, remote heart rate estimation, and face anti-spoofing.

返回