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

20180905-28 马锦华:Towards Assumption-free Unsupervised Domain Adaptation for . ...

2018-8-30 17:48| 发布者: 程一-计算所| 查看: 356| 评论: 0

摘要: 报告嘉宾:马锦华(中山大学)报告时间:2018年09月05日(星期三)晚上20:00(北京时间)报告题目:Towards Assumption-free Unsupervised Domain Adaptation for Visual recognition主持人:杨猛(中山大学)报告人 ...

报告嘉宾:马锦华中山大学

报告时间:2018年09月05日(星期三)晚上20:00(北京时间)

报告题目:Towards Assumption-free Unsupervised Domain Adaptation for Visual recognition

主持人:杨猛中山大学


报告人简介:

马锦华,博士毕业于香港浸会大学,曾于美国罗格斯大学与美国约翰霍普金斯大学(世界综合大学排名前20)进行博士后科研工作,现为中山大学数据科学与计算机学院副教授;他的研究兴趣包括机器学习(分类器/特征融合、迁移学习、半/弱监督学习等)、计算机视觉(智能视频监控、行为识别、行人再识别等)、医疗大数据分析(缺失数据处理、时间序列分析、肝病诊断预测模型等);他的研究成果发表于机器学习与计算机视觉领域的国际权威期刊(如IEEE TPAMI,IJCV,IEEE TIP等)、国际权威会议(如ICCV,CVPR,ECCV、MICCAI等)、医学领域的国际权威期刊(如Alimentary Pharmacology & Therapeutics,Critical Care Medicine等);他是SCI杂志Journal of Electronic Imaging的副编辑和多个国际权威期刊的审稿人(如IEEE TIP, IEEE TCYB, IEEE TIFS, PR等)。

个人主页:

http://isee.sysu.edu.cn/~majh/ 


报告摘要:

It is well-known that the performance of a classifier / detector trained on one dataset will be degraded when applying on another dataset. It is called the dataset - bias problem. Domain adaptation has been proved to be an effective approach to solve the dataset-bias problem. Without labels in the target domain, many existing unsupervised domain adaptation methods assume that conditional distributions in source and target domains are equal to each other, so that the joint distributions can be aligned by matching the marginal distributions. However, it may not be able to verify whether such assumption is valid in practice. In this talk, I will briefly review the basic principle of domain adaptation. Then, I will report our recent research works on unsupervised domain adaptation for person re-identification and object recognition without using the equal conditional distribution assumption.


参考文献:

[1] Cross-domain person re-identification using domain adaptation ranking SVMs, Andy J Ma, Jiawei Li, Pong C Yuen and Ping Li, IEEE Transactions on Image Processing (TIP), Vol. 24, No. 5, pp. 1599-1613, 2015.

[2] Semi-Supervised Region Metric Learning for Person Re-identification, Jiawei Li, Andy J Ma and Pong C Yuen, International Journal of Computer Vision (IJCV), In press, 2018.

[3] Learning domain-shared group-sparse representation for unsupervised domain adaptation, Baoyao Yang, Andy J Ma, and Pong C Yuen, Pattern Recognition 81: 615-632, 2018.

[4] Domain-Shared Group-Sparse Dictionary Learning for Unsupervised Domain Adaptation, Baoyao Yang, Andy J Ma, and Pong C Yuen, AAAI Conference on Artificial Intelligence, 2018.

[5] Dynamic Label Graph Matching for Unsupervised Video Re-identification, Mang Ye, Andy J Ma, Liang Zheng, Jiawei Li, and Pong C Yuen, International Conference on Computer Vision (ICCV), pp. 5152-5160. 2017.


18-28期VALSE在线学术报告参与方式:


长按或扫描下方二维码,关注“VALSE”微信公众号(valse_wechat),后台回复“28期”,获取直播地址。



特别鸣谢本次Webinar主要组织者:

VOOC责任委员:杨猛中山大学


活动参与方式:

1、VALSE Webinar活动依托在线直播平台进行,活动时讲者会上传PPT或共享屏幕,听众可以看到Slides,听到讲者的语音,并通过聊天功能与讲者交互;

2、为参加活动,请关注VALSE微信公众号:valse_wechat 或加入VALSE QQ群(目前A、B、C、D、E、F、G群已满,除讲者等嘉宾外,只能申请加入VALSE H群,群号:701662399);

*注:申请加入VALSE QQ群时需验证姓名、单位和身份,缺一不可。入群后,请实名,姓名身份单位。身份:学校及科研单位人员T;企业研发I;博士D;硕士M。

3、在活动开始前5分钟左右,讲者会开启直播,听众点击直播链接即可参加活动,支持安装Windows系统的电脑、MAC电脑、手机等设备;

4、活动过程中,请不要说无关话语,以免影响活动正常进行;

5、活动过程中,如出现听不到或看不到视频等问题,建议退出再重新进入,一般都能解决问题;

6、建议务必在速度较快的网络上参加活动,优先采用有线网络连接;

7、VALSE微信公众号会在每周一推送上一周Webinar报告的总结及视频(经讲者允许后),每周四发布下一周Webinar报告的通知及直播链接。


[slides]

最新评论

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

GMT+8, 2018-9-21 16:30 , Processed in 0.039031 second(s), 19 queries .

Powered by Discuz! X3.2

© 2001-2013 Comsenz Inc.

返回顶部