PROJECT TITLE :

Detect2Rank: Combining Object Detectors Using Learning to Rank

ABSTRACT:

Object detection is a vital research space in the sector of laptop vision. Several detection algorithms are proposed. But, each object detector relies on specific assumptions of the item look and imaging conditions. As a consequence, no algorithm will be thought of universal. With the big variety of object detectors, the subsequent question is how to pick and combine them. During this paper, we tend to propose a framework to find out how to combine object detectors. The proposed method uses (single) detectors like Deformable Part Models, Color Names and Ensemble of Exemplar-SVMs, and exploits their correlation by high-level contextual features to yield a combined detection list. Experiments on the PASCAL VOC07 and VOC10 knowledge sets show that the proposed method significantly outperforms single object detectors, DPM (eight.4%), CN (half-dozen.eight%) and EES (17.zero%) on VOC07 and DPM (vi.five%), CN (five.5%) and EES (16.2%) on VOC10. We have a tendency to show with an experiment that there are not any constraints on the kind of the detector. The proposed methodology outperforms (2.four%) the state-of-the-art object detector (RCNN) on VOC07 when Regions with Convolutional Neural Network is combined with different detectors employed in this paper.