Real-Time Object Detection using Deep Learning

S . Dhanalakshmi , Excel Engineering College (Autonomous) , Komarapalayam.

Abstract
Central the convolutional neural network is the convolutional layer that gives
Real time object detection is a vast, vibrant and sophisticated
the network its name. This layer performs an operation known as
area of computer vision aimed towards object identification and
“convolution”.
recognition. Object detection detects the semantic objects of a class
In the context of a convolutional neural network, a convolution may be a
objects using OpenCV (Open source Computer Vision), which is a
linear operation that involves the multiplication of a group of weights with the
library of programming functions mainly trained towards real time
input, very similar to a standard neural network. as long as the technique was
computer visionin digital images and videos. Visually challenged
designed for two-dimensional input, the multiplication is performed between
people cannot distinguish the objects around them. The main aim
an array of input file and a two-dimensional array of weights, called a filter or
behind this real time object detection is to help the blind to overcome
a kernel.
their difficulty. Real time object detection finds its uses in the areas
like tracking objects, video surveillance, pedestrian detection, people
The filter is smaller than the input file and therefore the before the sort of
counting, self-driving cars, face detection, ball tracking in sports and
multiplication applied between a filter-sized patch of the input and the filter
many more. This is achieved using Convolution Neural Networks,
may be a scalar product. A scalar product is that the element-wise
which is a representative tool of Deep learning. This project acts as an
multiplication between the filter-sized patch of the input and filter, which is
aiding tool for visually challenged people
then summed, always leading to one value. Because it leads to 1 value, the
operation is conventionally represented and mentioned because the
Keywords: Convolutional Neural Network, OpenCV, Deep Learning.
“scalar product”.

Using a filter smaller than the input is intentional because it allows an

I. INTRODUCTION equivalent filter (set of weights) to be multiplied by the input array multiple
Object detection is a technology to detect various objects in digital times at distinct points on the input. Specifically, the filter is applied
images and videos too. It is mainly helpful within the self- driving cars, systematically to every overlapping part or filter-sized patch of the input file,
face detection, etc., where the objects are to be continuously monitored. left to right, top to bottom.
The algorithm or the technique involved for object detection during this
project is Convolutional Neural Networks which is a class of Deep
learning. This uses MobileNet SSD technique during which MobileNetis
a neural network used for image classification and recognition whereas
SSD is a framework that is used to realize the multibox detector. The Input CN Output
mixture of both MobileNet and SSD can do object detection. The main image
advantage or purpose of choosing Deep learning is that we do not need
to do feature extraction from data as compared to machine learning. npiinn

The Haar-like trait play a crucial role in detecting the objects in a picture.
This systematic application of an equivalent filter across a picture may be a
They scan the entire picture starting from the top left and compares
powerful idea. If the filter is meant to detect a selected sort of feature within
every small box with the trained data. In this way, even small-detailed
the input, then the appliance of that filter systematically across the whole
objects present within the imagesare identified.
input image allows the filter a chance to get that feature anywhere within the
image.
II. METHODOLOGY This capability is usually represented and mentioned as translation invariance,
e.g. the total altogether concern in whether the feature is present instead of
Deep learning, a subset of machine learning which in turn is a subset of where it should had been present.
artificial intelligence (AI) has networks capable of learning things from
the data that is unstructured or unlabeled. The approach utilized in this
project is Convolutional Neural Networks (CNN). It uses the Haar-
cascade classifiers which help us in the detection of objects.

1. CNN:
The convolutional neural network, or CNN for brief, could also be a
specialized kind of neural network model designed for working with
two-dimensional image data, although they're going to be used with one-
dimensional and three-dimensional data.

Fig1.2. Image classification using CNN

2. OpenCV:
Open CV stands for open source computer vision. it's a group of libraries all the thing s are identified and every object is surrounded by an oblong box
in Python. it's a tool by which we will be able to manipulate the pictures , and therefore the name of the object is additionally displayed. we'll be only
like image scaling, etc. This supports and helps us in developing real observing the output video stream but not the input video stream.
time computing applications. It mainly concentrates and targets on image
processing, video capture and analysis. It includes several features like
face detection and also object detection. Currently OpenCV supports III. RESULT
differing types of programming languages like C++, Python, Java etc., Here, in this project we’ve considered around 15 to 20 objects to be detected
and it's available on various platforms including Windows, Linux, OS X, during the training. Some of those include ‘person’, ‘car’, ‘train’, ‘bird’,
Android ‘sofa’, ‘dog’, ‘’plant’, ‘aero plane’, ‘bicycle’, ‘bus’, ‘motorbike’, etc.
etc.
The output of this project displays the objects detected with a rectangular box
around the object with a label indicating it’s name and therefore the exactness
3. Training the data set:
with which the object has been detected on the top of it. It can dig out any
The data set is typically the gathering of knowledge . the info set could
number of objects existing during a single image with certainty
also be collection of images or alphabets or numbers or documents and
files too. the info set we used for the thingdetection is that the collection
of images of all the objects that are to be identified. Several different
images of every and each object is typically present within the data set.
If there are more number of images like each object within the datasets
then the accuracy are often improved. The important thing that's to be
remembered is that the info within the data set must be labelled. there'll
be actually 3 data set. they're the training data set, the validation dataset
and therefore the other one is testing data set. The training data set will
usually contains around 85-90% of the entire labelled data. This training
dataset are going to be training our machine and therefore
the model is obtained by training the info set. The validation data set
consists of around 5-10% of the entire labelled data.

4. Developing a real time object detector:

For developing a true time object detector using deep learning and open
cv we'd like to access our web cam during a really effective way then the
thing detection is to be applied to each and every frame. we should
always install open cv in our systems.The deep neural network module
should be installed. Firstly, we should always always import all the
specified packages:

1. From imutils.video we'll import VideoStream

2. From imutils.video we'll import FPS
3. we'll import numpy as np
4. we'll import argparse
5. we'll import imutils
6. we'll import time
7. we'll import cv2

The next step is to construct the argument parse then we should always
parse the arguments.
--prototxt: provide path to the Caffe prototxt file.
--model: provide path to the pre-trained model.
--confidence: The minimum probability threshold to filter weak
detections. The default value is given as 20%.
The next step is to initialize CLASS labels and corresponding random

COLORS.
Each object when it's detected, it's surrounded by a box with some
predefined colour. Thus, we assign each object a specific color.
After that we'll load our model and that we will provide the regard to our
prototxt and also to our model files. With the assistance of imutils we'll
read the video and that we will set the amount of frames per second.
Now with this some predefined number of frames are going to be loaded
per second. Eachframe is analogous to the image. Now these images are
going to be given because the inputs to the model. The model will
process the input image and produces the output image which consists of
labels. in additional practical sense the input raw image is given to the
model. Now the model process the input image. within the output image
APPLICATION
VI. REFERENCES
Here are a some of the future implementation of object detection. 1. Geethapriya S, N. Duraimurugan, S.P. Chokkalingam, “Real-Time Object
1. Face detections and recognition: Detection with Yolo”, International
Face detection perhaps be a separate class of object detection. We Journal of Engineering and Advanced Technology (IJEAT)
wonder how some applications like Facebook, Faceapp, etc., detect and 2. Abdul Vahab, Maruti S Naik, Prasanna G Raikar an Prasad S R4,
recognize our faces. this is often a sample example of object detection in “Applications of Object Detection System”, International Research Journal of
our day to day life. Face detection is already in use in our lifestyle to Engineering and Technology (IRJET)
unlock our mobile phones and for other security systems to scale back 3. Hammad Naeem, Jawad Ahmad and Muhammad Tayyab, “Real-Time
rate . Object Detection and Tracking”,IEEE
4. Meera M K, & Shajee Mohan B S. 2016, "Object recognition in images",
International Conference on InformationScience (ICIS).
2. Object tracking: 5. Astha Gautam, Anjana Kumari, Pankaj Singh: "The Concept of
Object detection is additionally utilized in tracking objects like tracking Object Recognition", International Journal of Advanced Research in
an individual and his actions, continuously Computer Science and Software Engineering, Volume 5, Issue 3,
March 2015
monitoring a ball within the game of Football or Cricket. As there's an 6. Joseph Redmon, Santosh Divvala, Ross Girshick, “You Only
enormous interest for people in these games, these tracking techniques Look Once: Unified, Real-Time Object Detection”, The IEEE
enables them to know it during a better way and obtain some additional Conference on Computer Vision and Pattern Recognition
information. Tracking of the ball is of maximal importance in any ball- (CVPR),2016,pp. 779-
based games to automatically record the movement of the ball and adjust 788
the video frame accordingly. 7. V. Gajjar, A. Gurnani and Y. Khandhediya, "Human Detection and
Tracking for Video Surveillance: A Cognitive Science Approach," in
3. Self-driving cars: 2017 IEEE International Conference on Computer Vision Workshops,
this is often one among the main evolutions of the planet and is that the
2017.
best example why we'd like object detection. so as for a car to travel to
the specified destination automatically with none human interference or
to form decisions whether to accelerate or to use brakes and to spot the
objects around it. this needs object detection.

4. Emotions detection:
this permits the system to spot the type of emotion the person puts on his
face. the corporate Apple has already tried to use this by detecting the
emotion of the user and converting it into a respective emoji within the
smart phone.

5. Biometric identification through retina scan:

Retina scan through iris code is one among the techniques utilized in
high security systems because it is one among the
foremost accurate and unique biometric.

6. Smart text search and text selection (Google lens)

In recent times, we've encountered an application in smart phones called
google lens. this will recognize the text and also images and search the
relevant information within the browser without much effort.

V. CONCLUSION
Deep-learning based object detection has been a search hotspot in recent
years. This project starts on generic object detection pipelines which
give base architectures for other related tasks. With the assistance of this
the 3 other common tasks, namely object detection, face detection and
pedestrian detection, are often accomplished. Authors accomplished this
by combing 2 things: Object detection with deep learning and OpenCV
and Efficient, threaded video streams with OpenCV. The camera sensor
noise and lightening condition can change the result because it can create
problem in recognizing the objects. generally, this whole process
requires GPU’s rather than CPU’s. But we’ve done using CPU’s and
executes in much less time, making it efficient. Object Detection
algorithms act as a mixture of both image classification and object
localization. It takes the given image as input and produces the output
having the bounding boxes adequate to the amount of objects present
within the image with the category label attached to every bounding box
at the highest. It projects the scenario of the bounding box up the shape
of position, height and width.