Localize Any Object in X-Ray Security Scans without Human Annotation
Self-supervised object localization in X-ray scans using synthesized training examples and granularity-aware supervision.
COMPUTER VISION · UNIVERSITY OF TRENTO
PhD Student MHUG
Hi, I’m Yaqi Cai, a first-year Ph.D. student at the University of Trento, supervised by Prof. Nicu Sebe. I joined MHUG in 2025 and currently collaborate with Elisa Ricci, Feng Xue, and Teng Long.
I received my M.Sc. degree from the DUT-RU International School of Information Science and Engineering, Dalian University of Technology, in 2024, advised by Assoc. Prof. Yu Liu and Assoc. Prof. Qi Jia, with collaboration with Prof. Nan Pu. I received my B.Eng. degree from the School of Software, Dalian University of Technology, in 2020, advised by Assoc. Prof. Jing Gao.
My research focuses on Novel Class Discovery and X-ray vision, especially object detection, segmentation, and 3D reconstruction.
RESEARCH
Recent and selected work. See Google Scholar for my publication profile.
Self-supervised object localization in X-ray scans using synthesized training examples and granularity-aware supervision.
Coordinating supervised and unsupervised gradients to improve the separation of known and novel visual categories.
Discovering highly similar visual categories with region-aligned features and semi-supervised proxy learning.
Positive-augmented contrastive learning for more effective representation learning and novel category discovery.