AMPHIBIANDETECTOR: ADAPTIVE COMPUTATION FOR DETECTION OF MOVING OBJECTS
D. A. Svitov1,2, S. A. Alyamkin1
1Expasoft LLC, Novosibirsk, Russia
2Institute of Automation and Electrometry, Siberian Branch, Russian Academy of Sciences, Novosibirsk, Russia
Keywords: convolutional neural networks, detection, computer vision, deep neural networks
Abstract
Convolutional neural networks (CNNs) allow reaching the highest accuracy for the task of object detection in images. The major challenges in further development of object detectors are false-positive detections. In some applications, it is of interest to detect only moving objects: face of a person approaching the intercom or car in overall traffic. In this paper, we propose an approach to object detection, which makes it possible to reduce the number of false-positive detections by processing only moving objects. The proposed approach is a modification of CNNs already trained for the object detection task and can be used to improve the accuracy of an existing system by applying minor changes to the algorithm. The efficiency of the proposed approach is demonstrated on the open dataset "CDNet2014 pedestrian."
|
