A Unified Approach for Detection, Classification and Segmentation |
 |
|
Abstract: The task of recognizing objects in realistic scenes essentially requires the coordination of all the three vision taks: (image) classification, (object) detection and (region) segmentation. To tackle the challenging dataset presented in PASCAL VOC 2008 challenge, we have used a unified framework to integrate information obtained from each process into the other. We begin with the highly successful appearance-based detector of Felzenszwalb et al. and augment it with various contextual cues (extracted from the given training images and other Internet images) and region-based cues (derived from image segmentation) to achieve superior detection, classification and segmentation results. |
Publications"A Unified Approach for Detection, Classification and Segmentation" |