I am a Lecturer (Assistant Professor) affiliated with the Vision, Learning, and Control (VLC) group in the School of Electronics and Computer Science (ECS) at the University of Southampton.

I specialize in computer vision and machine learning for artificial general intelligence. I am committed to teaching AI/machines to comprehend the world through multimodal data, enabling them to achieve animal-level, or human-level, or superhuman-level general intelligence. My current focus is on making machine learning models more capable and controllable to understand the physical world, through scaling and aligning compute, data, models, and tasks for deep learning. I am interested in exploring machine learning methodologies, including generative modeling, continual learning, and affective computing.


  • 10/2023, We have openings for PhD positions: 'Generative Modeling in Computer Vision' and 'Deepfake Detection with Continual Learning'. Note that these positions are open to students from UK and Horizon Europe qualifying countries (meeting the criteria for UK home-level fees) as well as other international students.
  • 02/2023, We are opening some positions for talents who are keen to publish papers, open source code, and make an impact.
  • 02/2023, One interesting paper "Freestyle Layout-to-Image Synthesis" is accepted as a highlight (10% of accepted papers, 2.5% of submissions) to CVPR 2023. This paper introduces freestyle layout-to-image synthesis with diffusion generative network (FreestyleNet). The proposed FreestyleNet produces realistic-looking and creative images like ‘a warehouse running on a railroad’ and 'a hornless unicorn sitting on a bench'. Thanks all the reviewers for valuable comments and constructive suggestions!
  • 01/2023, I join University of Southampton as a Lecturer (Assistant Professor) in Computer Science.
  • 11/2022, One paper "An Efficient Recurrent Adversarial Framework for Unsupervised Real-Time Video Enhancement" gets accepted to IJCV.
  • 11/2022, Two papers "Isolation and Impartial Aggregation: A Paradigm of Incremental Learning without Interference" and "Riemannian Local Mechanism for SPD Neural Networks" are accepted at AAAI 2023.
  • 10/2022, One paper "Multi-agent Actor-Critic with Time Dynamical Opponent Model" gets accepted to Neurocomputing.
  • 09/2022, One inspiring paper "S-Prompts Learning with Pre-trained Transformers: An Occam's Razor for Domain Incremental Learning" gets accepted to NeurIPS 2022. This paper proposes a rule-breaking continual learning paradigm (S-Prompts). It does not require the accumulation of the knowledge from previously learned domains. Instead, it merely learns the knowledge independently domain by domain, through simply prompting the state-of-the-art transformers.
  • 08/2022, One paper "A Continual Deepfake Detection Benchmark: Dataset, Methods, and Essentials" gets accepted to WACV 2023.
  • 06/2022, One paper "Trilevel Neural Architecture Search for Efficient Single Image Super-Resolution" is accepted by the CVPR-2022 NAS workshop.
  • 03/2022, One paper "Generative Flows with Invertible Attentions" is accepted by CVPR 2022, with surprisingly increased final review scores (2 strong accepts and 2 borderline accepts). Thanks all the reviewers for the valuable comments and suggestions! This paper introduces an interesting idea of Invertible Transformer-based Attentions to Generative Normalizing Flows.
  • 10/2021, One paper "Neural Architecture Search for Efficient Uncalibrated Deep Photometric Stereo" is accepted by WACV 2022.
  • 09/2021, I join SMU Singapore as a tenure-track Assistant Professor in Computer Science.
  • more news...