行知论坛197:Detect and Segment Objects with Image-Level Labels

时间:2018-12-03浏览:10设置


告时间:20181206日(周四)下午 3:00

报告地点:计算机学院4042

 

报告题目:Detect and Segment Objects with Image-Level Labels

人:Zequn Jie, senior researcher in Tencent AI Lab

人:许春燕         

           

报告简介:Object detection and semantic segmentation heavily rely on expensive fine-grained annotations. Leveraging only cheap image-level labels shows great potential in object detection and semantic segmentation. This talk will briefly introduce our recent works on object detection and semantic segmentation with only image-level annotations. Furthermore, we will also show that image-level labels as auxiliary guidance, can boost the current standard semantic segmentation frameworks when combined with pixel-level labels. 

 

报告人简介:Dr. Zequn Jie is a senior researcher in Tencent AI Lab. Prior to coming to Tencent, he was a postdoctoral research fellow in the department of electrical and computer engineering of National University of Singapore. He received the bachelor and Ph.D. degree from University of Science and Technology of China and National University of Singapore, respectively. His research interests mainly fall in the fundamental computer vision topics, e.g. supervised and weakly-supervised object detection, localization and semantic segmentation. He has served as a reviewer of several top conferences and journals, including T-PAMI, T-IP, CVPR, NIPS, ICML, ECCV, AAAI and ACM MM.


返回原图
/