|
|
||
|
|
Kihwan Kim Research Scientist at NVIDIA Research Ph.D. in Computer Science Georgia Institue of Technology, CoC/GVU/CPL Advisor : Dr. Irfan Essa Member of CPL and Graphics Group Contact : email) kihwan23 at cc.gatech.edu (nvresearch) kihwank at nvidia dot com 2701 San Tomas Expressway Santa Clara, CA 95050 |
|
|
I joined NVIDIA Research from Jan.2012, my new page is here, however, I will keep this page updated |
|
|
Main research |
|
|
Curriculum Vitae [PDF] (updated 03-03-2012) |
|
|
|
Prediction of ROI in Scenes with Camera Motions |
|
We use stochastic fields for predicting important future regions of interest as the scene evolves dynamically. We evaluate our approach on a variety of videos of team sports. We show that our approach can detect where to move the camera based on observations in the scene and compare the detected/predicted regions of interest to the camera motion as generated by actual camera operators - "Detecting Regions of Interest in Dynamic Scenes with Camera Motions" PDF The paper will appear in IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2012) IEEE CVPR 2012 : PROJECT HOMEPAGE -Collaborators :Dongreyol Lee and Dr. Irfan Essa |
|
|
|
Gaussian Process Regression Flow |
|
Modeling a trajectory as a continuous dense flow field from a sparse set of vector sequences using Gaussian Process Regression. Meanflows and confidences (from variances) allow for incrementally predicting possible paths and detecting anomalous events from online trajectories. We evaluate various types of video data having complete and incomplete trajectories. - "Gaussian Process Regression Flow for Analysis of Motion Trajectories" PDF The paper was published in IEEE International Conference on Computer Vision IEEE ICCV 2011 : PROJECT HOMEPAGE
-Collaborators :Dongreyol Lee and Dr. Irfan Essa |
|
|
Dynamic Scene Analysis using Motion Field |
|
|
Player actions and interactions in dynamic sports scenes are complex as they are driven by many factors, such as the short-term goals of the individual player, the overall team strategy, the rules of the sport, and the current context of the game. We show that such constrained multi-agent events can be analyzed, and even predicted, by estimating the global movements of all players in the scene at any time and used to predict play evolution. - "Motion Fields to Predict Play Evolution in Dynamic Sports Scenes" PDF IEEE CVPR 2010 : PROJECT HOMEPAGE -Collaborators:Matthias Grundmann, Dr. Ariel Shamir, Dr. Iain Matthews, Dr. Jessica Hodgins and Dr. Irfan Essa |
|
|
Player Localization Using Multiple Static Cameras |
|
|
Modeling and analysis for the problem of fusing corresponding players' positional information as finding minimum weight K-length cycles in complete K- partite graphs. We use our proposed algorithm-class for an end-to-end sports visualization framework, and demonstrate its robustness by presenting results over 60,000 frames of real soccer footage captured over five different illumination conditions, play types, and team attire. - "Player Localization using Multple Static Cameras for Sports Visualization" PDF IEEE CVPR 2010 : PROJECT HOMEPAGE -Collaborators: Raffay Hamid, Ram Krishan Kumar, Matthias Grundmann, Dr. Jessica Hodgins and Dr. Irfan Essa |
|
|
Augmenting Earth-Maps with Dynamic Information |
|
|
Augmented Earth Maps visualize the live broadcast of dynamic sceneries within a city. We propose different approaches to analyze videos of pedestrians and cars, under differing conditions and then augment Aerial Earth Maps (AEMs) with live and dynamic information. We also analyze natural phenomenon (clouds) and project information from these to the AEMs to add the visual reality.PROJECT HOMEPAGE - Journal of Virtual Reality Springer 2011 [PDF] - IEEE International Symposium on Mixed and Augmented Reality (ISMAR) 2009 [PDF](TBA) - Presentation [PPT] - Media coverage : CNN, New Scientist, Popular Science, Discovery Channel, Technology Review (MIT), Engadget, Vizworld, Revolution Magazine, etc. -Collaborator : Dr. Irfan Essa , Dr. Sangmin Oh and Jeonggyu Lee |
|
|
Real-time Transparent-Colored Shadow |
|
|
We provide a general non-manifold meshes and an additional extension to shadows of transparent casters. We first introduce a generalization of an object’s silhouette to non-manifold meshes. By using this generalization, we can compute the number of caster surfaces between the light and receiver, and furthermore, we can compute the light intensity arrived at the receiver fragments after the light has traveled through multiple colored transparent receiver surfaces.
- Journal of Graphics Tools(JGT) 2008 [PDF] - Technical Report in GT-IC-07-04 [PDF] - GT-CMU Graphics retreat 2007 [PDF] - Collaborator : Dr. Byungmoon Kim, Dr. Greg Turk |
|
|
GPS-Ray: Reconstruction of Urban Scene using GPS |
|
|
The main idea of this research is started from the assumption that the change of SNR in gps receiver at certain location could discriminate obstruction/un-obstruction structure. Using this evidence, we can localize and reconstruct building structures by only using off-the-shelf GPS receivers.
- IEEE ISWC 2008 : [PDF] - Presentation: [PPT] - Gatech Technical Report GT-IC-08-06 : [PDF] SNR test(Test 20070602) : and Heightmap/EM(Test 20070621) -Collaborator : Dr. Jay summet, Dr. Thad Starner , Dr.Irfan Essa |
|
|
|
Video based Non-Photorealistic Rendering |
|
Making Non-photorealistic Rendering(NPR) system using global gradient field from Radiail Basis interpolation and dispersion filters (water-colorization). For temporal coherence we adopt Michael Black's piecewise-smooth flow fields( robust regularization). Dispersion filter is also designed for mimicing pigment dispersion on the water fluid.
Project for Samsung STAR/SAIT 2008 -Collaborator : Dr. Irfan Essa |
|
|
|
Multi-scale Mosaic Generation for Video navigation |
|
Making mosaic using Labeled multi-scale tiling algorithm. The mosaic enables the users to navigate easily and remix the video scenes for their convenience. In matching process we used the annotated information from Family Video Archive's(Aware home) xml architecture. This work is presented at Living Game World Symposium 2006 and Also appeared in ACM Multimedia 2006.
- ACM International Conference on Multimedia 2006 [PDF] - Computational Photography project link -Collaborator : Dr. Irfan Essa, Dr. Gregory Abowd |
|
|
|
Face Recognition using GSVD |
|
Making Face recognition application with GSVD. we used Linear Discriminant Analysis with Generalized Singular Vector Decomposition which reduces dimension of input data image. ( k-1 dimension where k is number of class at training step ) Application is developed under OpenCV and Visual C++ environment.
- Face Recognition using LDA with generalized SVD -Collaborator : Sangmin Lee, Dr. James M. Rehg , Dr. Haesun Park |
|
|
|
Real-time Face Detection |
|
- Implementation of two well-known face detection algorithm. The link below contains the explanation of the algorithm,source code and executable binary files. Face Detection by Viola-Jones and Morphological
operator page link
If you want to use more reliable viola-Jones' module check this link Intel OpenCV's Viola-Jones face-detector link it is easy to use :) - Short cut link for demo movie |
|
|
|
|
|
Research and Development at Samsung SDS IT R&D Center |
|
|
|
|
|
|
Face Recognition System |
|
Samsung IT R&D Center made a 'Face Recognition System' in 2002 after 3 years of research . It was a proto-type system, named 'ViaFace'. Later, it was used in various field and industry including well known Korean apartment franchise 'Raemian' and Some Mexico airport etc. This System consists of 2 types, Verification (one to one), Identification (one to many or Survailance).
- More details.. - Presented at Comdex 2001 Las vegas
|
|
|
|
Real-time Collaboration System : Synbiz |
|
Syncbiz is a real-time collaboration system, which includes Application sharing module, Text chatting module, Video/Audio conferencing module, Shared virtual directory module, Multiuser White board module and Realtime Agenda Mgr(Scheduler) module. One session permits 10 concurrent users and all users share each modules. And One syncbiz local server has a capacity to sustain 50 concurrent sessions under main server which controls 50-capacity local server.
- More details.. - Samsung IT R&D Best eSolution award 2003 - Syncbiz introduction page in Romanian (Thanks to Alexandra Seremina in Novosibrirsk State Univ.) -Collaborator : Taesoo Jun, Yongho Woo and Joonsung Park |
|
|
|
IP-STB Framework : LivingWise CS |
|
LWCS is an framework for IP Set top box made by Samsung Electronics and KT (Korea Telecommunication). It managed overall I/O and controllers and has poweful applications built on Microsoft Windows CE environment. It contains MP3 and media player for IPSTB, Realtime news feed and weather forecasting by RSS.
- More details.. -Collaborator : Taesoo Jun and Joonsung Park |
|
|
|
|
|
Miscellaneous research |
|
|
|
|
|
|
Illumination Subspace Reconstructing images with arbitrary lighting condition using at least three various directions of light-source. [PDF] |
|
|
Building Recognition with SIFT features Building recognition by classifying features from Scale Invariant Feature(SIFT) Detection [PDF] |
|
|
Quick-time VR Making Quick-time VR movie using Sift ,Warping and RANSAC Algorithms. In every face, 25 image warping is applied. To make homography, RANSAC is applied to fitting the matrix. - Creating cubic VR Movie Part1 - Creating cubic VR Movie Part2 Homography - Creating cubic VR Movie Part3 : Automatic fitting by RANSAC Algorithm |
|
|
|
|
|
|
![[PDF]](http://www-static.cc.gatech.edu/~kihwan23/GTCMU/trans_shadow.JPG){kind=link}