Projects
Automated Macular Pathology Diagnosis in Retinal OCT Images , Georgia Tech, Summer 2009-now
-
This is a novel problem domain in the analysis of optical coherence tomography (OCT) images:
We design an effective descriptor for diagnosis in OCT images with a machine-learning based framework.
[MICCAI'10 paper, PDF] [Slides] [Detailed Project Webpage]
Named Entity Recognition from Spoken Documents, National Taiwan Univ., 2002-2004
-
We proposed to use an efficient data structure, the PAT trees, to extract global evidences of named entities from the whole spoken documents.
Significant performance improvements were obtained.
[IEEE ASRU'05 paper, PDF]
1st Place Award in Game design competition, First Chunghwa Cup National Computer Game Development Competition, Taiwan, 2001
-
Our team name is Peregrine, and our advisors are Prof. I-Peng Lin and Prof. Feipei Lai.
I am mainly in charge of designing the particle system that is used to simulate fire, snow, rain, and splendid magics. [Fancy Video Trailer (83M .avi)]
 |
 |
 |
Cosmos Exploration (Term Project of Computer Graphic), Fall 2001
-
We build up a fancy 3D cosmos system from the scratch that enables the players to fly through the cosmos, watching the beautiful planets and nebulas.
 |
 |
 |
 |
 |
 |
Tri-View Morphing (term project of Digital Image Processing, Prof. Hung) Spring 2003
-
Estimate the camera poses from correspondence points in three images. Generate novel views by warping and
blending the three input images according to the estimated camera poses and disparity maps.
 |
| Generate arbitrary novel view within the triangle plane in real-time |
Face Verification (term project of course: Pattern Recognition, Prof. Hung), Fall 2002
-
Verify the identity of a face image by using MDA and PCA with several image-preprocessing techniques.
 |
| examples of eigenface (PCA algorithm) |
2D Morphing (term project of Digital Image Processing, Prof. Pei) Fall 2001
-
Generate the intermediate image sequences from two input images.
 |
 |
| assign feature lines | generate intermediate image sequences from two input face images |