Yi Li
 

Ph.D student, ECE Dept, University of Maryland

College Park, MD 20742

liyi AT umiacs DOT umd DOT edu

Home
Curriculum vitae
Publications
Ongoing research

Past research  

Teaching

Beyond research


 


 

Education 

·  Aug 2004 – present

        Computer Vision Lab, University of Maryland, College Park

       Advised by Prof. Yiannis Aloimonos and Dr. Cornelia Fermuller

·   Aug-2001 - Oct 2004
       
M.Eng in Computer Science, (1st rank)

·   Aug-1998 - Oct 2001
B.Eng in Computer Science, (First Honor)

        South China University of Technology, Guangzhou

 
Research Interests 

·    Human movement analysis.

·    Visual perception: optical illusion, etc

·    Robotics 

·    Computer vision

·    Neuroscience: Mirror neuron,  


Awards and Honors 

·      Jan 2008,  Future Faculty Fellow, A. James Clark School of Engineering, University of Maryland. (for potential for securing a tenure-track faculty position at a top-50 engineering school.)

·      2nd place, 1st Semantic Robot Vision Challenge (sponsored by  NSF), AAAI2007, Vancouver, Canada, 2007.

·      Best student paper, 10th International Conference on Frontiers in Handwriting Recognition, 2006.

 

Ongoing Research (Research Statement)

1. Human Movement Analysis in Social Intelligence.

Social Signal Processing

Social intelligence is the facet of our cognitive abilities that guides us through the complex web of our everyday interactions, whether these require us to be a respected colleague on the workplace, a careful parent at home, a leader in our community, or simply a person others like to have around in a moment of relaxation.

 

Coordinated Action

Coordinated action is one of the basic abilities for social interactions. All animal species grouping for defensive, reproductive or hunting needs have evolved complex communicative behaviors to obtain coordinated action (Frith, 2008). Joint action in humans has been formalized in many ways and constitutes one of the most interesting and recent chapters of cognitive neuroscience (Sebanz, 2006). Yet the quantification of this ability remain somewhat elusive, especially outside controlled experimental settings. The main problem is related to the complexity of the topic due to the large space of parameters one might want to consider in a numerical analysis. Furthermore, the formal analyses of such issues require also the clear definition of some metrics to define the efficacy of inter-individual action coordination.

 

Parkinson and CDC

 

 

2. Active Robot Vision and Visual Perception

 

Vision Based Robot

Mobile robots must have the ability to navigate autonomously and recognize objects in a previously-unknown indoor environment in many tasks. However, state-of-the-art techniques in robotics and in computer vision are not used together to achieve this goal. In this paper, we present a complete system that segments objects and utilizes the online image datasets for object recognition during navigation. Equipped with a quad camera, the mobile platform explores an unknown indoor area using SLAM, and separates the object from the background actively. Shape-based matching is performed on a set of shape images and the detected objects are highlighted with a bounding box and semantic label

 

Visual perception of the intensity

 

The contrast effect and its counterpart, the assimilation effect, paradoxically co-exist in intensity perception. It is natural to assume that there is a single structural mechanism that accounts to both effects. However, despite there is a large body of work studying individual illusions, the link between them is not well understood. Here, we propose a computational model using L1 minimization. Analog to the Ganglion cell in the visual path, the intensity of the stimulus is compressed by simple local average filters at random positions. This process can be formulated using an underdetermined linear system. Assuming the signal is sparse in the Fourier domain, the decoder finds the sparse solution of this system in the reconstruction phase. Experiments show that the contrast and the assimilation effect are the artifacts during the signal reconstruction, which suggests that both effects are the two sides of the same coin

 


 

 

Past Research

 

3D silhouette consistency. CVPR 2007 (Joint work with Prof. David Jacobs)

Shape Recognition. BMVC 2007  (Joint work with Xiaodong Yu)

Text line segmentation. IEEE Trans on PAMI 2007  (Joint work with Dr. David Doermann, Dr. Yefeng Zheng, and Dr. Stefan Jaeger)

 

Language Identification. ECCV 2008, PR 2008  (Joint work with Guangyu Zhu)

Airline ticket recognition. IEEE Trans on PAMI 2004 (joint work with Zhiyan Wang)

 

 

 

Teaching (Teaching Philosophy)

CMSC 250

Co-Instructor

Discrete Structures (with Dr. Fawzi Emad)

This is the introduction to to the algorithm for undergraduates majoring in computer science. 70 students, 25% workload.

 

Last updated: Oct 2009

 Under Construction