Abstract
From a complementary perspective to model development, data-centric AI aims to improve and analyse data to better understand AI systems. In this talk, Dr Liang Zheng will introduce some attempts from my group at ANU. Specifically, he will talk about how to synthesize training data for applications where real data are sensitive or hard to obtain, such as face recognition and micro-expression recognition.
He will also discuss how to evaluate the difficulty of the test environments, or in other words, the model accuracy, in an unsupervised way. Finally, he will introduce a new video format from which motions can be efficiently captured by existing video processing architectures. He will conclude with perspectives of data-centric problems.
Speaker
Dr Liang Zheng (School of Computing, ANU) is an Associate Professor (tenured) in the Australian National University and holds a Future Fellowship by the Australian Research Council.
He is best known for his contributions in object re-identification and tracking, where he published well-known datasets and algorithms such as Market-1501 (ICCV 2015), random erasing data augmentation (AAAI 2020), part-based convolutional baseline (ECCV 2018), and joint detection and embedding (ECCV 2020). His recent research interest is data-centric computer vision.
He regularly serves as an Area Chair for CVPR, ICCV, ECCV, and NeurIPS, and a Program Co-Chair for ACM Multimedia 2024 and AVSS 2024. He received his B.S degree (2010) and Ph.D degree (2015) from Tsinghua University, China.