Application of Cascade Methods as a Universal Object Detection Tool

D. P. Matalova,b,*, S. A. Usilina,b,**, D. P. Nikolaevb,c,***, and V. V. Arlazarova,b,****

a Federal Research Center “Computer Science and Control” of the Russian Academy of Sciences, Moscow, 119333 Russian Federation

b Smart Engines Service LLC, Moscow, 121205 Russian Federation

c Institute for Information Transmission Problems of the Russian Academy of Sciences, Moscow, 127051 Russian Federation

Correspondence to: * e-mail: d.matalov@smartengines.com
Correspondence to: ** e-mail: usilin@smartengines.com
Correspondence to: *** e-mail: dimonstr@iitp.ru
Correspondence to: **** e-mail: vva@smartengines.com

Received 17 October, 2022

Abstract—This paper is devoted to a review of the achievements of the Moscow scientific school of image recognition, formed under the leadership of Professor Vladimir L’vovich Arlazarov, in the field of development and application of the Viola–Jones method. One of the main areas of research at the school is the development of computationally efficient recognition algorithms, which requires a deep understanding of the problem and a wide expertise in the field of existing classical algorithms. Such classic method as the Viola—Jones method became an essential tool to solve a wide range of image recognition problems. This paper provides an overview of the modifications of the original method developed by the scientific school and describes in detail the experience of solving many different practical problems that arise in the development of modern energy-efficient image recognition systems.

Keywords: machine learning, Viola–Jones method, scientific school, image processing, edge computing, object detection, image classification, image analysis, statistical recognition methods

DOI: 10.1134/S1054661823040302