I received my Ph.D. degree at Huazhong University of Science and Technology in Dec. 2022, advised by Prof. Xinge You and worked closely Prof. Ling Shao.
My current research interests span computer vision and machine learning with a series of topics, such as zero-shot learning, generative modeling and learning, and visual-and-language learning.
News
2024.09, Our paper on audio-visual zeroshot learning is accepted to IEEE TETCI, congrating to Yujie.
2024.08, Our paper on domain generalization is accepted to IEEE TMM, congrating to Guanglin.
2024.07, Our TWO papers on zero/few-shot learning is accepted to ACM MM'24, congrating to Shuhuang and Yunwei.
2024.05, Our paper on few-shot object detection is accepted to IEEE TIP, congrating to Zhimeng.
2024.02, Our TWO paper on zero-shot learning are accepted to CVPR'24.
2024.02, Our SURVEY on few-shot object detection is accepted to Information Fusion, congrating to Zhimeng.
2024.01, Our paper on Non-Transferable Representation Learning is accepted to ICLR'24 (Spotlight), congrating to Ziming.
2023.12, Our paper on zero-shot learning is accepted to Science China Information Sciences, congrating to Shuhuang.
2023.11, I give a talk in Harbin Institute of Technology (Shenzhen), invited by Prof. Haijun Zhang.
2023.09, I give a talk in ZHEJIANG LAB, invited by Dr. Jin Zhao.
2023.08, one paper on zero-shot learning is accepted to TEC.
2023.07, Our paper on rainy image generation is accepted to ICCV'23.
2023.07, I am invited as an Area Chair (AC) of PRCV'23.
2023.06, I give a talk in University of Science and Technology of China, invited by Prof. Hongtao Xie.
2023.04, Our paper on zero-shot learning is accepted to ICML'23.
2023.04, I give a talk in Alibaba DAMO Academic, invited by Baigui Sun.
2023.04, I give a talk in Huazhong Agricultural Univeristy, invited by Prof. Hong Chen.
2023.04, I give a talk in Guizhou University, invited by Prof. Yisong Wang.
2023.03, I am invited as an Area Chair (AC) of VALSE, news at Here.
2023.03, I give a talk in National Key Laboratory of Science and Technology on Multispectral Information Processing, invited by Prof. Yi Chang.
2022.12, Our TransZero++ is accepted to TPAMI.
2022.08, I was invited as a Program Committee (PC) Member for AAAI'23.
2022.05, I give a talk about our CVPR'22 work (MSDN) in VALSE.
2022.05, I give a talk about zero-shot learning in AI TIME and AI Drive.
2022.04, Our paper on zero-shot learning is accepted to IJCAI'22.
2022.04, We have released the full codes of TransZero accepted to AAAI'22.
2022.03, Our paper on zero-shot learning is accepted to CVPR'22.
2022.02, I gave a talk in Extreme Mart (极市).
2022.02, Our paper on zero-shot learning is accepted to TNNLS.
2022.02, I gave a talk in AI TIME PhD-NeurIPS, invite by AI TIME.
2022.01, I start a Research Intern at Tencent AI Lab.
2021.12, Our paper on zero-shot learning is accepted to AAAI'22.
2021.09, Our paper on zero-shot learning is accepted to NeurIPS'21.
2021.07, Our paper on zero-shot learning is accepted to ICCV'21.
Researches
A key challenge of artificial intelligence is to generalize machine learning models from seen data to unseen scenarios.
Zero-shot learning (ZSL) is a typical research topic targeting this goal. ZSL aims to classify the images of unseen classes by constructing a mapping relationship between the semantic and visual domains.
Although ZSL has achieved significant progress, there have a numbers of essential challenges. Recently, large-scale VLM-based ZSL method is popular, e.g., CLIP, it is an extension of the classical ZSL.
Dr. Shiming Chen has been focusing on tackling bottleneck challenges to promote ZSL (especially for the classical ZSL), covering fundamental questions of
How to enhance the visual features by alleviating the cross-dataset bias between pre-train dataset and ZSL benchmarks?
How to discover the intrinsic semantic knowledge by alleviating the visual-semantic domain shift problem?
How to align the visual and semantic features in a common space by reducing the discrepancy between the heterogeneous visual-semantic representations?
Specifically, his three representatives research projects are:
1. Developing the visual feature enhancement algorithms to tackle the challenge of cross-dataset bias in ZSL.
As for the embedding-based ZSL, a graph-guided dual attention network is introduced to fuse the local visual features and explicit global visual features
to enhance visual features. As for the generative ZSL, several feature refinement learning methods are proposed to enhance the visual features and encourage
the generator to synthesize realistic visual features for unseen classes. The papers of this project have been published in
ICCV'21,
IJCAI'22,
IEEE TNNLS'22,
etc.
2. Developing the effective ZSL algorithms to tackle the visual-semantic domain shift problem.
As for the embedding-based ZSL, a attribute-guided Transformer network and mutually semantic distillation network are proposed to learn the intrinsic
semantic knowledge, enriching the visual features with semantic information to enable desirable semantic knowledge transfer from seen calsses to unseen ones.
As for the generative ZSL, dynamic semantic prototype learning is proposed to refine the pre-defined semantic prototypes under the guidance of visual signal,
aligning the empirical and actual semantic prototypes for synthesizing accurate visual features. The papers of this project have been published in
CVPR'22,
AAAI'22,
IEEE TPAMI'22,
IEEE TEC'23,
ICML'23 ,
CVPR'24, etc.
3. Developing the semantic-visual adaptation framework for visual-semantic alignment.
Different to existing one-step adaptation method that on alignment the feature distributions between visual and semantic domains,
this method utilizes a hierarchical adaptation to learn an intrinsic common space for semantic and visual feature representations
by adopting sequential structure adaptation and distribution adaptation. The papers of this project have been published in
NeurIPS'21,
CVPR'24.