Pico "El Aguila" (4118 m). Venezuela (2002)

 

 

Rómer E. Rosales, Ph.D.

Postdoctoral Fellow
Probabilistic and Statistical Inference Group

Artificial Intelligence Group
romer[at]psi.toronto.edu

Phone: 416-946-8810
Office: BA 4159


Electrical and Computer Engineering,

Computer Science

University of Toronto
Toronto, Ontario
Canada


I moved to the Boston area, and I am at the MIT CS-AI Laboratory (CSAIL)

 

My newer web page

 

Research Interests


 

Development of novel machine learning algorithms and architectures

Approximate statistical inference
Probabilistic data clustering and graph inference.

Representation and probabilistic models in computer vision and image processing
3D articulated body pose estimation from single monocular images

Graphics and visualization of complex data

Data mining and bioinformatics

 

Publications  


Rómer Rosales, Kannan Achan, and Brendan Frey
Learning to Cluster using Local Neighborhood Structure
(pdf|ps)

To appear in Proc. International Conference on Machine Learning (ICML).  Jul 2004.

 

Rómer Rosales, Kannan Achan, and Brendan Frey
Unsupervised Image Translation
(pdf|ps)

In Proc. International Conference on Computer Vision (ICCV).  Oct 2003.

Example translations

 

Rómer Rosales, Kannan Achan, and Brendan Frey
Translating Images by Unsupervised Estimation of Switching Filters
(pdf|ps)

In Proc. IEEE Workshop on Statistical Signal Processing  (SSP) (invited paper).  Sep 2003.

 

 

Rómer Rosales and Stan Sclaroff
Combining generative and discriminative models for inferring articulated body pose from a single image (AKA The Specialized Mappings)
(pdf|ps)

Submitted (03/03).

Example applications

 

Rómer Rosales and Brendan Frey
Learning Generative Models of Affinity Matrices
(pdf|ps)

In Proc. 19th Conference on Uncertainty in Artificial Intelligence (UAI). Aug 2003.

More examples on new data (soon)

 

Rómer Rosales and Stan Sclaroff
Algorithms for Inference in the SMA

In Proc. 5th IEEE International Conference on Automatic Face and Gesture Recognition (FG2002). Presented in FG2002, Washington, DC, May 2002.

 

Rómer Rosales
The Specialized Mappings Architecture with Applications to Vision-Based Estimation of Articulated Body Pose

Ph.D. Thesis. Jan 2002. (pdf|ps|Abstract)

 

Rómer Rosales and Stan Sclaroff

Learning Body Pose Via Specialized Maps

In Proc. Neural Information Processing Systems NIPS-14, 2002. Presented at NIPS, Vancouver, BC, Dec 2001.

 

 

Rómer Rosales, Matheen Siddiqui, Joni Alon, and Stan Sclaroff

3D Body Pose through Virtual Cameras
(or the boring-title version: Estimating 3D Body Pose using Uncalibrated Cameras)

In Proc. IEEE Computer Vision and Pattern Recognition (CVPR). Presented at CVPR, Kauai, Hawaii, Dec 2001.

 

Rómer Rosales, Vassilis Athitsos, and Stan Sclaroff

3D Hand Pose Reconstruction Using Specialized Mappings

In Proc. IEEE International Conference on Computer Vision (ICCV). Presented at ICCV, Vancouver, BC, Canada, Jul 2001.

 

Rómer Rosales and Stan Sclaroff

Specialized Mappings and the Estimation of Human Body Pose from a Single Image

In Proc. IEEE Workshop on Human Motion (HUMO). Presented at HUMO, Austin, TX, Dec 2000.

 

Rómer Rosales and Stan Sclaroff

Inferring Body Pose without Tracking Body Parts

In Proc. IEEE Computer Vision and Pattern Recogniton (CVPR). Presented at CVPR, Hilton Head Island, SC, Jun 2000.

 

Rómer Rosales and Stan Sclaroff

Learning and Synthesizing Human Body Pose and Motion

In Proc. 4th IEEE International Conference on Automatic Face and Gesture Recognition (FG2000). Presented at FG200, Grenoble, France, 2000.

 

Rómer Rosales and Stan Sclaroff

Trajectory Guided Recognition

In Proc. SPIE 99. Presented at SPIE, Boston, MA, 1999.

 

Rómer Rosales and Stan Sclaroff

A Framework for Heading-Guided Recognition of Human Activity (pdf| ps)

To appear in Computer Vision and Image Understanding (CVIU)  2003.

 

Rómer Rosales and Stan Sclaroff

3D Trajectory Recovery for Tracking Multiple Objects and Trajectory Guided Recognition of Actions.

In Proc. IEEE Computer Vision and Pattern Recognition (CVPR). Presented at CVPR, Fort Collins, CO, 1999

 

Rómer Rosales

Recognition of Human Action Using Moment-Based Features.

Boston University Computer Science Technical Report BU 1998-020, Boston, MA, 1998.

 

Rómer Rosales and Stan Sclaroff

Improved Tracking of Multiple Humans with Trajectory Prediction and Occlusion Modeling.

In Proceedings IEEE CVPR Workshop on the Interpretation of Visual Motion. Presented at CVPR,  Santa Barbara, CA, 1998

Past Presentations On-line

 

Generative Models of Affinity Matrices (html|ps 4 s/p)

Rómer Rosales and Brendan Frey
Presented at NIPS Workshop on Spectral Methods in Dimensionality Reduction, Clustering, and Classification. Vancouver, BC, Dec 2002.

   


Other Links

Latex thesis format (due to high demand)