David Liu at CMU

            About AMP Lab        Projects        Downloads        Publications        People        Links

Group Member



David Liu

PhD. Candidate

Personal Homepage: http://dliu.cmu.googlepages.com

Email: dliu@cmu.edu

Office: PH B23
Lab: Porter Hall B6
Phone: 412-268-1171 (Office)

Fax: 412-268-1679

Mailing Address:
Porter Hall B23, CMU ECE

5000 Forbes Avenue

Pittsburgh, PA 15213

[Research Interests]        [Project]      [Publications

Research Interests

  • Currently working on Object Detection/Recognition [Download papers (CVPR_BP) (ICASSP) (ICIP)] 

My text detection system received the most accurate results in a recent VACE evaluation. 

My recent interest is on Markov Random Field (MRF), probabilistic Latent Semantic Analysis (PLSA) and the Semantic-Shift algorithm.

Presentation slides on PLSA

Presentation slides on Semantic-Shift

Video demos accompanying the Semantic-Shift presentation slides [ 1 ]  [ 2 ]  [ 3 ]

  • Shape Matching based on Shape Context and ICP

The shape context method can be integrated into the iterative closest point (ICP) method as an auxiliary feature vector, enriching the representation of an image point from spatial information only, to spatial and shape information.  This yields a registration method more robust than the original ICP method.  The method is general for 2D shapes.  It does not calculate derivatives, hence being able to handle shapes with junctions and discontinuities.


Final translation error  6.3 pixels

Final rotation error  7.5 degrees

ICP with soft shape context

Final translation error  1.6 pixels

Final rotation error  1.8 degrees


Final translation error  8.2 pixels

Final rotation error  6.1 degrees

ICP with soft shape context

Final translation error  0.2 pixels

Final rotation error  5.7 degrees


  • Probabilistic Relevance Feedback for Image Retrieval

    More can be found at the probabilistic relevance feedback link.



  Top of this page



  • Content-based Image Retrieval

For details on the project please see the trademark retrieval project and the probabilistic relevance feedback project.

  • Object Detection and Recognition

The discovery website is up.

Top of this page



Journal Papers: 

      D. Liu and T. Chen, "DISCOV: A Framework for Discovering Objects in Video", IEEE Transactions on Multimedia, 2008.

      C.M. Huang, D. Liu and Li-Chen Fu, "Visual Tracking in Cluttered Environment Using the Visual Probabilistic Data Association Filter", IEEE Transactions on Robotics, 2006.

Conference Papers: 

Top of this page

This website is maintained by Devi Parikh
Copyright 2005 Advanced Multimedia Processing Lab. All rights reserved.
Revised: Jan 1, 2006.