设为首页收藏本站

VALSE

VALSE 首页 活动通知 好文作者面授招 查看内容

20160824-28 杨建伟:Unsupervised Representation Learning Jointly with...

2016-8-20 13:08| 发布者: 程一-计算所| 查看: 6111| 评论: 0

摘要: 报告嘉宾1:杨建伟(弗吉尼亚理工学院)报告时间:2016年8月24日(星期三)上午10:00(北京时间)报告题目:Unsupervised Representation Learning Jointly with Image Clustering主持人:庄连生(中国科学技术大学 ...

报告嘉宾1:杨建伟(弗吉尼亚理工学院)

报告时间:2016年8月24日(星期三)上午10:00(北京时间)

报告题目:Unsupervised Representation Learning Jointly with Image Clustering

主持人:庄连生(中国科学技术大学)



报告摘要:

Unsupervised learning from visual data has drawn much attention in the past few years. A bunch of methods have been proposed to learn deep representations from unlabeled image data. In this talk, I will introduce our method about unsupervised representation learning jointly with image clustering. Intuitively, meaningful image clusters can provide supervisory signal to deep representation learning, while good representations can help to get meaningful clusters. We proposed to learn deep representation and cluster unlabeled images in a progressive and iterative manner. In our framework, successive operations in a clustering algorithm are expressed as steps in a recurrent process, stacked on top of representations output by a Convolutional Neural Network (CNN). During training, image clusters and representations are updated jointly: image clustering is conducted in the forward pass, while representation learning in the backward pass. By integrating two processes into a single model with a unified weighted triplet loss and optimizing it end-to-end, we can obtain not only more powerful representations, but also more precise image clusters. Extensive experiments show that our method outperforms the state-of-the-art on image clustering across a variety of image datasets. Meanwhile, the learned representations generalize well when transferred to other tasks.


参考文献:

[1] Jianwei Yang, Devi Parikh, Dhruv Batra. Joint Unsupervised Learning of Deep Representations and Image Clusters. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016.


报告人简介:

Jianwei Yang is now a Ph.D. student at Virginia Tech, under the supervision of Dr. Devi Parikh. Prior to that, he received his M.E. degree from BUPT jointly with CBSR & NLPR, CASIA, advised by Prof. Stan Z. Li, and mentored by Prof. Zhen Lei. During that period, his work on face anti-spoofing won the first place at the 2nd Competition on Counter Measures to 2D Face Spoofing Attacks. He got his B.E. degree from Central South University, Changsha. His recent research interests fall in unsupervised learning from visual data and vision/language modeling.


[Slides]

最新评论

Archiver|手机版|小黑屋|Vision And Learning SEminar    

GMT+8, 2020-9-26 23:32 , Processed in 0.030105 second(s), 19 queries .

Powered by Discuz! X3.2

© 2001-2013 Comsenz Inc.

返回顶部