This work proposes to learn visual encodings of attention patterns that enables sequential attention for object detection in real world environments. The system embeds a saccadic d...
Inspired by psychophysical studies of the human cognitive abilities we propose a novel aspect and a method for performance evaluation of contour based shape recognition algorithms ...
Abstract. This paper introduces a new framework for image classification using local visual descriptors. The pipeline first performs a nonlinear feature transformation on descripto...
A new, combined human activity detection method is proposed. Our method is based on Efros et al.'s motion descriptors[2] and Ke et al.'s event detectors[3]. Since both m...
Josef Kittler, Seiji Ishikawa, Takehito Ogata, Wil...
Many visual search and matching systems represent images using sparse sets of "visual words": descriptors that have been quantized by assignment to the best-matching symb...