REDUCTION OF FEATURE DESCRIPTION DIMENSIONALITY IN THE PROBLEM OF OBJECT DETECTION IN IMAGES
Lobachevsky State University of Nizhniy Novgorod; student
Read the full article
Abstract. A modification of sliding-window algorithm for object detection based on feature selection with the use of decision trees ensemble is proposed. Reduction in detection time without a loss in quality is demonstrated by the example of digital simulation of HOG-features in the problem of pedestrian detection.
Keywords:
object detection, pedestrian detection, histograms of oriented gradients, feature selection, decision tree
