Benjamin S. Riggan
Date of this Version
K. Nikhal, "Learning Discriminative and Efficient Attention for Person Re-Identification Using Agglomerative Clustering Frameworks", M.S. thesis, Department of ECEN, University of Nebraska–Lincoln, 2021
Recent advancements like multiple contextual analysis, attention mechanisms, distance-aware optimization, and multi-task guidance have been widely used for supervised person re-identification (ReID), but the implementation and effects of such methods in unsupervised person ReID frameworks are non-trivial and unclear, respectively. Moreover, with increasing size and complexity of image- and video-based ReID datasets, manual or semi-automated annotation procedures for supervised ReID are becoming labor intensive and cost prohibitive, which is undesirable especially considering the likelihood of annotation errors increase with scale/complexity of data collections. Therefore, this thesis proposes a new iterative clustering framework that incorporates (a) two attention architectures that learn to ignore background clutter and focus on important regions in the image, (b) three objective functions that produce discriminative feature representations without using any labels and (c) a diversity term that helps cluster persons across different cross-camera views without leveraging any identification or camera labels. Our approach provides new state-of-the-art performance on both image- and video-based datasets while reducing the performance gap between supervised and unsupervised ReID.
Advisor: Benjamin S. Riggan