Poster presentation, ICML, Vienna, Austria
This is a remote talk on the ICML 2021 conference. In this work, we propose a principled framework to learn both the local-instance and global-semantic structure of graph representations in a self-supervised way.
Poster presentation, ECCV, Glasgow, Scotland, UK
This is a remote talk on the ECCV 2020 conference. In this work, we propose a Learning to Combine framework, in which the knowledges acquired from multiple source domains are aggregated and applied to the target domain.
Oral presentation, CVPR, Washington, USA
This is a remote talk on the CVPR 2020 conference. In this work, we propose a Graph-induced Prototype Alignment (GPA) framework to align source and target domain for cross-domain detection tasks.
Oral presentation, AAAI, New York, USA
This is a remote talk on the AAAI 2020 conference. In this work, we propose to promote adversarial domain adaptation with both pixel-level and feature-level domain mixup.