I am excited to share the great news that I am going to formally join the Department of Computer Science at City University of Hong Kong as a Tenure-Track Assistant Professor in 2024 Fall. My research team to be built is Statistics & Machine Learning (SMILE) Labortary.

For the  academic year of 2024~2025, we have several openings for Ph.D. students and Postdocs, and also welcome various forms of collaboration. Please refer to Labortary and Join Us for detailed information regarding research direction and application instruction. Please feel free to reach out to me at <[email protected]> if you are interested.

The application instruction in Chinese can be found in wechat, and I appreciate the support from PaperWeekly [Link].

Our CVPR 2024 paper MoDE: CLIP Data Experts via Clustering is out. The Full version of the paper and code has been publihed [Link].

I feel great honor to be selected by CVPR Doctoral Consortium, and I am looking forward to seeing you in June.

 

I am currently offering 1~2 positions for research projects regarding multi-modal learning toward visual/multimedia understanding. In particular, we will play with the latest large language model and rethink the vision-based tasks including image understanding and video understanding. A 3-point research course will be credited for each semester and each position. The projects will start in the semester of Summer 2023. One Senior Researcher will participate, and we will give suggestions jointly.

Preferred background/experience: PyTorch, Detectron2, experience with LLM, Fundamental understanding of {Optimization, Transformer & CNN, Word Embedding}

We have two papers accepted by CVPR 2023 for presentation. More details including PDF, Supp and Code will be updated under the tab Research.

DiGeo: Discriminative Geometry-Aware Learning forGeneralized Few-Shot Object Detection

Supervised Masked Knowledge Distillation for Few-Shot Transformers

I am thrilled to share the good news that my ICLR paper is accepted.

TempCLR: Temporal Alignment Representation with Contrastive Learning

In this work, we work on video representation learning and proposed a contrastive learning framework to model the temporal succession between video and demonstrations.

More details including PDF, Supp and Code will be updated under the tab Research.

 

Looking for RA to work on a research project of visual attribute discovery starting in August. A 3-point research course will be credited for FALL 2022 for each position and the work done in the summer will also be counted. RA is expected to work at least 8 hours per week, meet with PhD mentors every week to discuss progress, and demonstrate clear contributions by the end of the semester.

Please reach out to <[email protected]> and <[email protected]> if you are interested.

Preferred Background & Technique:

• • Familiar with coding in Pytorch.
  • Experience and knowledge in video processing and clustering algorithms.

We have two papers accepted by CVPR 2022 for presentation. More details including PDF, Supp and Code will be updated under the tab Research.

Task-Adaptive Negative Class Envision for Few-Shot Open-Set Recognition

Few-Shot Object Detection with Fully Cross-Transformer

I am currently offering two Research Assistant positions for projects regarding few-shot learning towards visual/multimedia understanding. A 3-point research course will be credited for each semester and each position. The projects will start in the semester FALL 2021. Another Ph.D. student will take part in the projects and we will give suggestions jointly.

Please send me your CV to my email ([email protected]) if you are interested. I will schedule an interview starting from late Aug and will try my best to reply to each of your email within 24 hours.

We have two papers accepted by ICCV 2021 for poster presentation. More details including PDF, Supp and Code will be updated under the tab Research.

Partner-Assisted Learning for Few-Shot Image Classification

Query Adaptive Few-Shot Object Detection with Hierarchical Graph Convolutional Networks