Prof. Yi Yang

Sydney University of Science and Technology

时间：2019年9月19日10:00-11:00

地点：浙江大学玉泉校区曹光彪楼东楼502

本次讲座计读书报告一次

主讲人简介：

    杨易，悉尼科技大学人工智能中心教授。2010年浙江大学获得计算机博士学位，2011年-2013年在卡内基梅隆大学作博士后研究，后在昆士兰大学担任DECRA高级研究员，2014年年底起在悉尼科技大学先后担任高级讲师、副教授，并于2017年5月晋升为教授。2018年11月，加入百度研究院顾问委员会。担任多个顶级会议的领域主席和高级程序委员。是IEEE TIP和IEEE TCSVT的编委。谷歌引用一万余次。曾获得全国优秀博士论文奖、澳大利亚研究理事会发青年研究奖、浙江省科学奖一等奖、谷歌教师研究奖、澳大利亚计算机学会金质创新奖等。

Abstract：

    We mainly focus on the design of deep neural networks to enable the agents with a human-like visual understanding of the world. To leverage video temporal dynamics for complex action recognition, we studied 3D convolutional neural networks for more efficient video classification. We also worked on the interactions between human and object in constrained environments, e.g., kitchen. These videos are ego-centric, containing subtle motion changes. Beyond recognition, we worked on accurate video localization for large-scale clip-level retrieval, which can be applied in many real-world applications, e.g., online video search systems. A more challenging task would be predicting the future from past frames. It requires more sophisticated reasoning and intelligence. To accurately predict future frames, we designed new sequence network for better spatio-temporal modelling. Notably, we won many international competitions in video analysis.