Skip to main content


Model for multiclass object localization

Research Achievements

Model for multiclass object localization

IGERT trainee Carolina Galleguillos, with IGERT advisor Serge Belongie, and Professor Brian McFee and Professor Gert Lanckriet, introduced a novel model for multiclass object localization that incorporates different levels of contextual interactions (at the pixel, region and object level). The framework learns a single similarity metric from multiple kernels, combining pixel and region interactions with appearance features, and then uses a conditional random field to incorporate object level interactions. The interdisciplinary collaboration resulted in a new approach for kernel combination for nearest neighbor classification. It was recently published in CVPR 2010 and Carolina received an honorable mention award for her presentation on this work at the UCSD Jacobs School of Engineering 29th Research Expo.