Visual Artificial Intelligence
Visual Artificial Intelligence
Visual Artificial Intelligence
Visual Artificial Intelligence

About Us

Generalized Category Discovery:
A new and realistic open-world problem.
SeSDF: Self-evolved Signed Distance Field for Implicit 3D Clothed Human Reconstruction.
Learning Attention as Disentangler
for Compositional Zero-shot Learning
DreamAvatar: Text-and-Shape Guided 3D Human Avatar Generation via Diffusion Models
Open-Vocabulary Semantic Segmentation
with Decoupled One-Pass Network

Visual AI Lab (VAIL) is a research group directed by Dr. Kai Han, working on computer vision, machine learning, and artificial intelligence, under Department of Statistics and Actuarial Science, The University of Hong Kong. The overarching goal of our research is to achieve principled and comprehensive visual understanding, close the intelligence gap between machines and humans, and build reliable AI systems for open-world use. Our current research focuses on open-world learning, 3D vision, generative AI, foundation models and their relevant fields.

🚩 Openings:
(1) PhD students: We are always looking for strong students to work on exciting research problems (☞ fellowships and scholarships).
(2) Postdocs: Positions in computer vision and deep learning are available, with a competitive salary.
(3) Interns: HKU Summer Research Programme (☞ details), with the possiblity of conditional PhD offer with HKU-PS scholarship.
Please drop Dr. Kai Han an email with your resume if you are interested in working with us.

News and Updates

July 2023:

Two papers (on generalized category discovery/open-vocabulary semantic segmentation) are accepted to ICCV 2023.
Feb 2023:

Two papers (on compositional zero-shot learning/3D human digitization) are accepted to CVPR 2023.
Jul 2022:

One paper on novel category discovery without forgetting is accepted to ECCV 2022.
Jun 2022:

Best Paper Runner-Up Award at CVPR 2022 Workshop on Continual Learning in Computer Vision.
Mar 2022:

Three papers (about generalized category discovery/3D human reconstruction/instance segmentation) are accepted to CVPR 2022.
Jan 2022:

One paper about open-set recognition is accepted to ICLR 2022.
Oct 2021:

One paper about visual correspondence is accepted to BMVC 2021.
Sep 2021:

One paper about novel category discovery is accepted to NeurIPS 2021.