Real Time Object Dectection using machine learning

REAL TIME OBJECT DETECTION
Presented by:- Project Guide:-
Pratik Pratyay(12150052) Dr. Latha R Nair

• Our goal is to design a system that runs real-time object detection.Formally, given a
continuous camera live stream we want to successfully recognize(draw bounding box and
label the object) the moving object in a small amount of time.

• Monitoring cameras are used almost everywhere, and are producing immense video stream
everyday.Unfortunately, only a small portion of these data are interesting to the users.Let’s take
traffics surveillance as example: only the part of videos that actually has moving human or
vehicles are useful, therefore, recording everything all day long would inevitably become a
waste of storage, and would make going through video a miserable work for searching and
tracking.
• To address this problem, object detection system could be of great importance as it can detect
the objects with a high level of accuracy.
• In fact, the need of object detection algorithm to be applied on resource constrained
device is highly in demand and not limited to monitor camera (e.g. phones, sports watch).

• Several similar object detecting systems have been developed but their accuracies
were very poor.One such system is Object detection using DPM(Deformable part
model).
• Moreover their processing speed is also extremely less(around 0.07fps).

• The process of object detection is done by analyzing the input image and determining
the number of locations, sizes, positions of the objects.
• Object detection is the base for object tracking and object recognition, whose results
directly affect the process and accuracy of object recognition.
• A common object detection method is color-based approach which detects objects
based on their color values.

INPUT
• Capturing image from the computer webcam in real time environment.
OUTPUT
• Recognize the object and display its name along with probability.

SYSTEM:- Pentium IV 2.4 GHZ
HARD DISK:- 40GB
MONITOR:- 15 VGA Color
MOUSE:- Logitech
RAM:- 512 Mb

Environment Configuration
Before getting started with the coding part, we have to configure some
software to work properly with each other.
Required Softwares-
1.TensorFlow(python lib)/OpenCV
2. MinGW
3.Cmake
4.Sublime text editor

• TensorFlow is an open source software library for numerical
computation using data flow graphs. Nodes in the graph represent
mathematical operations, while the graph edges represent the
multidimensional data arrays (tensors) communicated between them.

• OpenCV is the most popular library for computer vision. Originally written in C/C++, it
now provides bindings for Python.
• OpenCV uses machine learning algorithms to search for faces within a picture. For
something as complicated as a face, there isn’t one simple test that will tell you if it
found a face or not. Instead, there are thousands of small patterns/features that must
be matched. The algorithms break the task of identifying the face into thousands of
smaller, bite-sized tasks, each of which is easy to solve. These tasks are also called
classifiers.
Refrence of papers:-
http://www.allresearchjournal.com/archives/2015/vol1issue9/PartG/1-9-20.pdf

The Whole process done by in two phases:
1. Training Phase
2. Object Identification Phase

Data set
Feature Extration
and classification
Training using algorithms
like CNN & SVM
Object Database

• The training phase makes use of a hybrid model, integrating the synergy of two superior
classifiers: Convolutional Neural Network (CNN) and Support Vector Machine (SVM), which
have proven results in recognizing different types of patterns.
• In this model, CNN works as a trainable feature extractor and SVM performs as a recognizer.
This hybrid model automatically extracts features from the raw images and generates the
predictions.
• Experiments have been conducted on the well-known MNIST digit database have shown that
compared to other methods the above fusion method has achieved better results. A recognition
rate of 99.81% without rejection, and a recognition rate of 94.40% with 5.60% rejection coul
• According to COCO data set(released by Microsoft)The whole 300k images is classified into 90
class
• In this phase we are going to train our system by using the 300k images.
• We are going to store all the feature and images into the object database

Capture Image
from Webcam
Background
Substraction
Extracting the
Feature using BCS
& FCS
Match the Features
with classes using
hybrid model
Display the name of
the object and its
probability
Object
Database

• In OIP an image can be detected on the basis of its hue, saturation and color value.
• Extract the feature from image by using BCS (Boundary Contour System)and FCS(Feature
Contour System).
• Match the extracted features with defined class by using a Hybrid model of CNN and SVM.
• Based on the extracted features, display the result by competitive learning(means the closest
match).

• In machine learning, a convolutional neural network (CNN, or ConvNet) is a class of deep, feed-
forward artificial neural networks that has successfully been applied to analyzing visual
imagery.
• CNNs use a variation of multilayer perceptrons designed to require minimal preprocessing.They
are also known as shift invariant or space invariant artificial neural networks (SIANN), based on
their shared-weights architecture and translation invariance characteristics.
• CNNs use relatively little pre-processing compared to other image classification algorithms.
This means that the network learns the filters that in traditional algorithms were hand-
engineered. This independence from prior knowledge and human effort in feature design is a
major advantage.

●
A Support Vector Machine (SVM) performs classification by finding the hyperplane that maximizes the margin between
the two classes
●
●
Notice: linearly separable and binary
It draws the widest channel, or street, between the two classes
The two class labels are +1 (positive examples) and -1 (negative examples)

• Microsoft classify all types of objects in 90 classes, containing 30k images. This Data
Set is known as COCO Data Set.
• CNN extract the feature from these data set and SVM classify the input according to
class.
All The data set resources link are give below:
• COCO DATA SET:
http://academictorrents.com/details/f993c01f3c268b5d57219a38f8ec73ee7524421a

CONT..
• Numerous face image sets are available on the web FERET face data set:
http://www.itl.nist.gov/iad/humanid/feret/
• UMIST data set:
http://images.ee.umist.ac.uk/danny/database.html
• Yale data set:
http://cvc.yale.edu/projects/yalefacesB/yalefacesB.html
• AR data set:
http://cobweb.ecn.purdue.edu/%7Ealeix/aleix_face_DB.html
• CMU PIE data set:
http://www.ri.cmu.edu/projects/project_418.html

• Face recognition may also be implemented and we can try to generate a security alert
system that will give alerts when an intruder is going to trespass any unauthenticated
area based on his/her face recognition. But all authenticated people face imprints data
has to be provided to the system in advance.
• These systems can also be used to detect suspicious objects such as weapons in
places like country borders.
• Monitoring cameras are used almost everywhere, and are producing immense video
stream everyday.Instead of that we can use this system.

1. Nidhi Kumari ,ISSN Print: 2394-750,Journal IJAR 2015;pg 1-9,396-399, 02-06-2015 Dept of
computer science NIT KURUKSHETRA ,Haryana ,INDIA. https//allresearchjournal.com
2. Khusboo khurana,Reetu avasthi Journal IJARCET ISSN: 2278-1323 April 4th 2015 Page 4-
13,232-241 SFS College Nagpur. https//ijarcet.org

• The concept is well extendable in applications like Intelligent Robots,Automatic Guided
Vehicles.
• Enhancement of Security Systems to detect the suspicious behaviour along with detection of
weapons.
• Identify the suspicious movements of enemies on boarders with the help of night vision
cameras and many such applications.
• In the proposed method, background subtraction technique has been used that is simple and
fast. This technique is applicable where there is no movement of camera. For robotic
application or automated vehicle assistance system, due to the movement of camera,
backgrounds are continuously changing leading to implementation of some different
segmentation techniques.

1. Viraktamath SV, Mukund Katti, Aditya Khatawkar,Pavan Kulkarni, “Face Detection and
Tracking using OpenCV,” The SIJ Transaction on Computer Networks & Communication
Engineering (CNCE), 2013, 1(3).
2. Pant A, Arora A, Kumar S, Arora RP. “Sophisticated Image Encryption Using OpenCV,”
International Journal of Advances Research in Computer Science and Software Engineering
3. Kevin Hughes – One more robot learn to see (http://kevinhughes.ca)
4. Belongie S, Malik J, Puzicha J. “Shape Matching and Object Recognition using shape contexts,”
IEEE Transactions on Pattern Analysis and Machine Intelligence, 2002
5. Tobias OJ, Seara R. “Image Segmentation by Histogram Thresholding Using Fuzzy Sets,” IEEE
Transactions on Image Processing, 2002; 11(12):1457-1465.
6. http://www.opencv.org

Real Time Object Dectection using machine learning

More Related Content

What's hot

Similar to Real Time Object Dectection using machine learning

Recently uploaded

Real Time Object Dectection using machine learning