THE 3 rd INTERNATIONAL SCIENTIFIC CONFERENCES OF STUDENTS AND YOUNG RESEARCHERS dedicated to the 99
th
anniversary of the National Leader of Azerbaijan Heydar Aliyev
175
REAL-TIME OBJECT DETECTION AND COUNTING Nazrin Gurbanova, Agamirza Fataliyev, Hasan Isgandarli Baku Higher Oil School Baku, Azerbaijan nazrin.gurbanova.std@bhos.edu.az agamirza.fataliyev.std@bhos.edu.az hasan.isgandarli.std@bhos.edu.az Supervisor: PhD, Associate Professor Leyla Muradkhanli Keywords: Python, OpenCV, Live-detection, Directions, Frame
With this thesis, a program written in the Python programming language,
using OpenCV library and working on the basis of Single-Shot Detection
(SSD) via implementing a convolution neural network (CNN) architecture, in
other words, will be described.
The reason for choosing the Python language as the main language is
its simplicity of syntax and versatility, which gives it an advantage over other
languages. The Single-Shot Detection allows us to detect multiple objects
within the image from one single shot, which in terms is the main reason we
divide the input/live video into frames and analyze each.
The utilization of the live-detection became more necessary in recent
years, as, for example, the spread of the coronavirus and the global
pandemic are forcing more shopkeepers, malls, restaurants, and other public
establishments to monitor the number of people in the premises and take the
necessary measures. The use of the above program makes it easily
accessible – the only things needed are the surveillance or any other camera
connected to the computer and the software. The usage of the software lets
the user select individual settings and update the code to add other features
later on. As an example, the alert system is included here as well, which is
triggered and sends the notification to the owner whenever the number of
people inside exceeds the threshold input.
Coming to the concept of how the
directions of the objects are identified, the
principle of calculating the difference of the
corresponding coordinates of centroids of the
objects is used and later compared to 0. On the
right the representation of the frame is provided
on which the objects are moving: the person
goes to the right, the car goes to the left, the dog
goes downwards, and the cat goes upwards. In
the following section the explanation of the
direction detection will be given: