Research Group of Prof. Dr. C. Sminchisescu
Institute for Numerical Simulation
maximize

Lic.-Ing. João Carreira

Address: Institut für Numerische Simulation
Wegelerstr. 4 (Flachbau)
53115 Bonn
Germany
Office: We4 0.021
Phone: +49 228 73 4091
E-Mail: carreira.ins.uni-bonn.de

Bio

I'm a doctoral student at the University of Bonn since January 2008, under the supervision of Prof. Cristian Sminchisescu. Previously, I completed a Licenciatura in Electrotechnical and Computer Engineering in the University of Coimbra, Portugal. Afterwards I did research on human computer interfaces using computer vision, also in Coimbra, supervised by Prof. Paulo Peixoto.

My CV is here.

Research

My goal is to develop models that are capable of automatically segmenting and recognizing objects present in an image.

Towards this goal, I'm studying a few specific problems: low-level figure-ground segmentation (which pixels project from a same object in the world), assignment (relative depth ordering and identification of boundaries that are occlusion artifacts) and recognition (relating new image measurements to previously seen exemplars).

I believe it's possible to generate segmentations that are sufficiently good without using shape priors. Sufficiently good means here that they allow for reliable and accurate recognition at some coarse but useful level (say at the level of "person", but not necessarily "person" + full pose).

Please visit the publications section of this website for pdfs of the papers below.

News: Paper accepted for oral presentation at ECCV 2012. It proposes a new region descriptor that makes it possible to obtain 47% VOC2011 segmentation score using efficient linear classifiers.
Check back in late September for source code.

Publications

J. Carreira, R. Caseiro, J. Batista and C. Sminchisescu. Semantic Segmentation with Second-Order Pooling. [Oral]
European Conference on Computer Vision (ECCV), October 2012

J. Carreira and C. Sminchisescu. CPMC: Automatic Object Segmentation Using Constrained Parametric Min-Cuts.
Accepted for publication in PAMI.

J. Carreira*, F. Li* and C. Sminchisescu. Object Recognition by Sequential Figure-Ground Ranking.
Accepted for publication in IJCV. (* first two authors contributed equally)

A. Ion, J. Carreira and C. Sminchisescu. Probabilistic Joint Image Segmentation and Labeling.
Advances in Neural Information Processing Systems (NIPS), December 2011.

A. Ion, J. Carreira and C. Sminchisescu. Image Segmentation by Figure-Ground Composition into Maximal Cliques.
In IEEE International Conference on Computer Vision (ICCV), November 2011.

J. Carreira, A. Ion and C. Sminchisescu. Image Segmentation by Discounted Cumulative Ranking on Maximal Cliques.
Technical Report 06-2010 (arXiv:1009.4823), Computer Vision and Machine Learning Group, Institute for Numerical Simulation, University of Bonn, June 2010.

J. Carreira and C. Sminchisescu. Constrained Parametric Min-Cuts for Automatic Object Segmentation.
In IEEE International Conference on Computer Vision and Pattern Recognition (CVPR), June 2010.

F. Li*, J. Carreira* and C. Sminchisescu. Object Recognition as Ranking Holistic Figure-Ground Hypotheses.
In IEEE International Conference on Computer Vision and Pattern Recognition (CVPR), June 2010. (* first two authors contributed equally)

J. Carreira, F. Li and C. Sminchisescu. Object Recognition by Ranking Figure-Ground Hypotheses.
Snowbird Learning, April 2010.

Code

Software for the CPMC algorithm is available here