Lei Chen（陈 磊）

I'm currently a Assistant Researchor of Department of Automation, Tsinghua University, China. Before that, I was an Associate Professor with the School of Computer and Communication Engineering, University of Science and Technology Beijing, China (Nov. 2022 to Jan. 2024). And I was a postdoctoral researcher of Department of Automation, Tsinghua University, China (Jun. 2020 to Oct. 2022). My current research interests lie in computer vision and pattern recognition.

I received the B.S. degree in the Qiushi Honers college, and the Ph.D degree with the school of electrical and information engineering, Tianjin University, China, in 2013 and 2020. I finished my PhD research at Tsinghua University advised by Prof. Jie Zhou, Prof. Jiwen Lu and Prof. Zhanjie Song.

Recent Selected Publications

(*Equal Contribution, #Corresponding Author)

We propose an ambiguousness-aware state evolution (AASE) method which represents the uncertainty of the input sequence and evolves the subsequent skeletons to generate a reasonable full-length sequence for action prediction.

We propose a weakly-supervised approach called Order-Constrained Representation Learn-ing (OCRL) to predict future actions from instructional videos by observing incomplete steps of actions.

We propose a recurrent semantic preserving generation (RSPG) method for action prediction. Our method learns to capture the tendency of observed sequences and complement the subsequent action with adversarial learning under some constrains, which preserves the consistency between the generation sequence and the observed sequence.

We propose an unsupervised representation method to learn principal orientations and residual descriptor (PORD) for action recognition. Our PORD aims to learn the statistic principal orientations and to represent the local features of action videos with residual values.

We propose a framework to enable existing methods to accommodate multiple data distributions by orthogonal subspaces. We introduce constraints on multiple hypersphere manifolds via MultiCenter Loss and employ a Spatial Split Strategy to ensure the orthogonality of base vectors associated with different hypersphere manifolds, corresponding to distinct distribution.

We propose a part-activated deep reinforcement learning (PA-DRL) method for action prediction. We design the PA-DRL to exploit the structure of the human body by extracting skeleton proposals under a deep reinforcement learning framework.

We propose an uncertainty-aware representation Learning (UARL) method for action segmentation. we design the UARL to exploit the transitional expression between two action periods by uncertainty learning.

We propose Skip-Plan, a condensed action space learning method for procedure planning in instructional videos. we abstract the procedure planning problem as a mathematical chain model.

We propose a prompt-based framework, Bridge-Prompt (Br-Prompt), to model the semantics across adjacent actions, so that it simultaneously exploits both out-of-context and contextual information from a series of ordinal actions in instructional videos.

We propose a deep embedding with discriminative sampling policy (DE-DSP) learning framework by simultaneously training two models: a deep sampler network that learns effective sampling strategies, and a feature embedding that maps samples to the feature space.

We propose a hierarchical equalization loss (HEL) by reconsidering the underlying factors affecting sample weights.

We propose a Multi-Scale Video-Text Correspondence Learning (MVTCL) framework, which enhances the grounding performance in complex scenes by modeling multi-scale semantic correspondence both within and between modalities.

We propose a Label Release and Contribution Degree Distinction (LRCDD)-based transfer subspace learning strategy to enhance recognition performance to solve these problems.

Academic Services

• Guest Editor: JVCI
• Conference Reviewer: CVPR, ICCV, ECCV, AAAI and so on
• Journal Reviewer: TIP, TMM, TCSVT and so on