讲座时间：2023年10月18日（星期四）15:30-16:30

讲座地点：人工智能与数据科学学院（西教一）102室  学术报告厅

讲座题目：面向无监督场景的行人重识别算法研究

讲座嘉宾：蓝龙  教授

摘要：无监督行人重识别在近些年来受到了广泛关注。该任务旨在利用无标签数据集训练模型，使得训练后的模型可以精确地判别不同摄像头下的相同行人。现阶段的无监督行人重识别方法往往依靠聚类算法来对数据集生成伪标签，然而这种方式生成的伪标签往往包含大量标签噪声，严重影响了模型的性能。我们的工作主要探索在标签噪声存在的情况下，如何使模型学习到更具辨别性的特征表达，从而针对行人取得更加准确的检索结果，目前我们的方法相比现有方法在该任务上取得了较大提升。

Long Lan is an Associate Professor of College of Computer Science and Technology at National University of Defense Technology. He was a Visiting PhD Student at University of Technology, Sydney (2015-2017). He received his Ph.D. degree in Computer Science from National University of Defense Technology (2013- 2017). His research interests lie in computer vision and machine learning, especially multi-object tracking, transfer learning and few shot learning. He has served as senior program committee of IJCAI’21, and published 60+ journal articles and conference papers, such as IJCV, TIP, TMM, TKDE, TCSVT, NeurIPS, ICML, ICLR, MM, KDD, AAAI, IJCAI, ICCV, ECCV.