Wenze Hu

PhD student, Department of Statistics, UCLA




Wenze Hu ()


Ph.D. Candidate Department of Statistics
University of California, Los Angeles
Advisor: Song-Chun Zhu


8145 Math Sciences Building
UCLA
Los Angeles, CA 90095

wzhuATstatDOTuclaDOTedu

Follows are my research projects:

Image Representation by Active Curves

File 0011.png, original image.
Corner saliency map.
Illustration of detected corners.Angles spanned by the corner are from red arm to green arm. For clarity, the arms only cover small portions of the curve templates of the corners.
Detected corner prototypes. Angles spanned by the corner are from red arm to green arm. For clarity, the arms only cover small portions of the curve templates of the corners.
Corner and curve prototypes. Angles spanned by the corner are from red arm to green arm. For clarity, the arms only cover small portions of the curve templates of the corners .
Deformed sketches from corners and arcs. Red dots denote corner centers .

A biologically inspired algorithm is implemented in this project, to capture the corners and long curves in natural images. [more]




Published in ICCV 2011.

Object Representation for View Invariant Recognition

Given a set of object images from various views, what is the correct representation so that computer can learn/memorize them in a coding efficient manner and recognize them in a computationally efficient manner? [more]



Published in CVPR 2010.

Below are projects I've done before:

Interactive 3D Scene Structure Annotation from Video

We designed a very easy procedure together with the software to interactively annotate and estimate the 3D structure of a scene.

Published in ICPR 2008.

Background Modeling integrating Primal Sketch and 3D Context.

This project proposed to combine both image color, texture and structure components to help discriminate foreground pixels as well as maintain a good scene background. (Left) A scene. (Middle) Segmenting the scene into regions of different property. (black: flat;red:texture;white:sketch) (Right) Results of foreground by incrementally applying different features on different image regions.

Published in CVPR 2008.

Digital Yuanmingyuan Project

I am only a member of a large team working on this project. For details of this project, please check http://www.otislab.cn/