( remove "hello" if you are human being )
|
Jiajun Zhu left UVa in Jan. 2008 and he is now working on computer vision problems which help making better maps. He spent a great time with his friends and colleagues at UVa from 2005 to 2007, where he was a member of the Computer Graphics group and his advisor was Prof. Greg Humphreys. Jiajun received his B. S in Computer Science from Fudan University in 2005. His research interests are Computer Vision, Computer Graphics, Computer Music, Human Computer Interaction, and a little bit Machine Learning. |
2007 |
Precomputed Radiance Transfer for Real-time Indirect Lighting using a Spectral Mesh Basis (PDF | Video |HTML) |
| Rui Wang, Jiajun Zhu, Greg Humphreys | |
| To appear in Proc. of Eurographics Symposium on Rendering (EGSR07), Grenoble, June 25-27, 2007 |
|
|
Fast Omni-directional 3D Scene Acquisition with an Array of Stereo Cameras (PDF) |
| Jiajun Zhu, Greg Humphreys, David Koller, Skip Steuart, Rui Wang | |
| To appear in Proc. IEEE International Conference on 3-D Digital Imaging and Modeling (3DIM07), Montréal, Canada, August 21-23, 2007. |
2005 |
Perceptual Visualization of A Music Collection (PDF | Video) |
| Jiajun Zhu, and Lie Lu | |
| Proc. IEEE International Conference on Multimedia and Expo (ICME05), Amsterdam, The Netherlands, June 6-8, 2005 |
2004 |
Musical Genre Classification by Instrumental Features (PDF) |
| Jiajun Zhu, Xiangyang Xue, and Hong Lu | |
| Proc. International Computer Music Conference(ICMC04), Miami, United States, Nov 1-5, 2004 |
Thesis |
An Implementation of Speech Recognition System based on Developmental Approach (Chinese version, PDF) |
| Jiajun Zhu | |
| Bachelor Honor Thesis, July, 2005 |
| 2005 - 2007: Computer Vision and Graphics | |
 |
Fast Omnidirectional 3D Scene Acquisition with an Array of Stereo Cameras: We present an omnidirectional 3D acquisition system based on a mobile array of high-resolution digital cameras. The stereo pairs are augmented with a time-varying lighting pattern created using standard photographic flashes, lenses, and patterned slides. The resulting system compares favorably with other state of the art 3D acquisition technologies in the resolution and quality of its output, and can be faster and less expensive than 3D laser scanners for digitizing large 3D scenes such as building interiors. (PDF) |
 |
Precomputed Radiance Transfer for Real-Time Indirect Lighting using A Spectral Mesh Basis: We present a novel method, based on precomputed radiance transfer, for rendering physically based, multi-bounce indirect lighting in real-time. Our key idea is to represent both the direct lighting and precomputed diffuse indirect transfer using a spectral mesh basis set derived from an arbitrary scene model. we demonstrate high-quality indirect lighting effects rendered at 15 30 fps with dynamically changing lighting and materials. (PDF | Video |HTML) |
 |
Structure and Motion from Unordered Image Collection: Inspired by the excellent previous work Photo Tourism and "How do I organize my holiday snaps?", I am currently trying to build a similar structure from motion system and improve the performance. For some preliminary results, visit the project webpage HTML. Here is the slides (.ppt) I made for UVa graphics seminar (may help to give you a quick review of the basic idea behind this). (project in progress, more details to come)
|
 |
High Dynamic Range Time Lapse Stereo: by capturing the natural illumination change and high dynamic range images, we are designing a robust stereo 3D reconstruction system based on the space-time stereo framework. The system is also designed to handle huge amount of data as our typical time lapse data set has more than 5,000 images (6GB storage). (Download Project Report: PDF)
|
 |
Structured Lighting: I spent 2 weeks implementing a simple structured lighting system with one projector and one video camera. The way I calibrate the projector-camera system might be slightly different from other proposed ones. I first measure the pixel-to-pixel mapping between projector and camera images (done automatically during the scanning process), then this mapping is used to generate checkerboard images from the projector viewpoint given the camera checkerboard images. The system is kind of slow (takes 40 seconds for one scan), but the reconstructed results are not too bad. (Download some reconstructed scenes in .ply format: scene1.ply, scene2.ply, scene3.ply) |
 |
Secure Interactive Rendering: We are looking at ways to secure graphics content in interactive applications. As models, textures, and shaders become important pieces of intellectual property, it is increasingly important that content creators be able to distribute manipulable versions of their 3D data while still maintaining some control on how those data can be copied or changed. We are examining perceptual approaches for securing 3D content without relying on high-latency remote rendering. Right now, several perceptual defense approaches are being investigated against computer vision based attacks such as shape from silhouette and shape from shading. |
| 2003 - 2004: Computer Music and Content-based Information Retrieval |
 |
eMusic: In this project, we tried to integrate many music content-based analysis components, such as tempo, rhythm, genre, and mood detection, play list generation, and music 2D/3D visualization in this system. Our goal is to help user better manage, browse, and search their personal music data. (PDF | Video)
|
 |
Music Genre Classification : Automatic musical genre classification is very useful for many musical applications. In this project, the features of instrument distribution and instrument-based notes are proposed to represent the high-level characteristics of music. Experimental results show that the proposed features have a good performance in musical genre classification. Comparison between our proposed features with the commonly used features - MFCC and MFCC with energy term illustrates that our proposed features perform better in discriminating some musical genres, such as Pop, Jazz, and Rock. (PDF) |
 |
eBand.Net : Have you ever dreamed of being a composer? eBand.Net allows you to create your own music with your friends through several very simple and interesting ways. For example, you can hum, whistle, and even paint the melodies. eBand.net will automatically combine all the melodies created by you and your friends. What's more? Your music can get published at the eBand.Net community. This is a team project and I am the project manager. We have 6 team members. This project lasts for 6 months and finally wins us the second place of Microsoft Imagine Cup 2005 software design competition in China. The awards include: 6 PDA, 6 Packages of Visual Studio.Net, and FREE trip to Brazil! |
 |
ePiano: ePiano can transcript simple polyphonic music. In another word, it can automatically convert wave file (wav) into Midi sound file. Moreover, the Virtual 3D Piano will show you how to play the music by piano. This is the first individual project in my sophomore year. It establishes my research interests in the field of audio processing and pattern recognition. ePiano was published as a shareware in July, 2004 and it has already been downloaded for more than ten thousands times. (Download Software ZIP)
|
 |
Video Game
Engine:
Building a video game from
scratch is the course project for
Prof. Luebke's
Real-time rendering and Game Technology class. Inspired by
Will Wright's
"Spore", Jerry, Kevin,
and I finally made an universe navigation type game.
(Download the final footage video
High|Low
| Visit individual webpage
HTML1,
HTML2, HTML3,
HTML4 |
Watch this video at Google Video:
Watch) |
 |
Recovering 3D face by a USB camera and a Lamp: In this course project, we implement an easy-to-build, low-cost, efficient but high-quality 3D face reconstruction system based on photometric stereo algorithm which can recover face model from three pictures taken at a fixed view point but with different lightings. It is extremely easy to use because neither camera nor lighting needs to be calibrated manually once the system is initially setup, and the system only consists of a USB camera and a desktop lamp. However, the accuracy of this system still needs to be improved. (Visit webpage HTML) |
 |
Evolving Virtual Creatures: This is the final course project for Intro Graphics course taught by Prof. Humphreys. In this project, Jerry and I tried to reproduce the famous work "Evolved Virtual Creatures" by Karl Sims in 1994. We finally came up with some creatures that can walk, swim, jump, or follow an object. It is pretty cool, check it out! (Visit webpage HTML | Download Video High|Low| Watch this video at Google Video: Watch)
|
 |
Ray Tracer: This is a course assignment in Intro Graphics course taught by Prof. Humphreys. I now realize making a ray tracer from scratch is not only doable but also very2 worth doing. In this project, I learnt many concepts such as geometry, lighting, color, shade, material, and a little modeling (by practice!). I implemented soft shadow, depth of field, motion blur, texture, super-sampling, and grid-based acceleration structures. The only bad thing is that I cannot really focus on other course-projects during that week. You know, compared with "branch prediction simulator", this one is much more exciting to me! (Visit webpage HTML) |
