CVPR Daily - Tuesday

Jingya Wang is a final year PhD student at Queen Mary University of London. She speaks to us ahead of her poster presentation today. Jingya tells us that her work is about jointly learning the attribute and identity of person re-identification under an unsupervised setting . Matching the person from non- overlapping camera views in different locations. Most existing person re- identification works under a supervised setting that needs a large amount of human annotation. Her work focuses on the transferable unsupervised setting, transferring knowledge from the existing dataset to the unseen new dataset. The work focuses on the jointly- modelled global identity and the local attribute information, because there is a heterogeneous problem for multi- task learning, so she proposes a progressive knowledge fusion mechanism by encoder-decoder networks for intermediate space that progresses transfer for the domain adaptation. “The reason we transfer the knowledge in the attribute space is because normally for the human re- identification, the ID labels from different domains, different data sets, are independent – they don’t have overlaps – but in our study, we want to transfer the knowledge. ” 12 Jingya Wang Tuesday