Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

OBJECT CLASSIFICATION

Submitted in partial fulfilment of the requirements for the award of degree of


BACHELOR OF ENGINEERING
IN
COMPUTER SCIENCE & ENGINEERING

Submitted to:
ASS. PROF. CHARN PREET KAUR

Submitted By:
GAURAV DOONGARWAL
(18BCS1824)
UTKARSH SHRIVASTAVA
(18BCS2271)

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING


Chandigarh University, Gharuan
September 2020
1. Introduction:
This project (Object Recognition) is about building a software program in which we
are able to recognise a given object and classify it using python. This project will help
us in recognising the particular objects with the help of some data set. This
application works for free and provide us quick responses. Due to its cost, this
application is very useful for the business purpose.

2. Project Design:

2.1. Methodology/Procedure: -
This project is built on the python platform and this application will able to run on the
all Desktop Operating Systems. We will build this project with the help of the Jupyter
Notebook and a programming language i.e. Python. The IDE i.e. Jupyter Notebook
provides us the environment by which we can write our code for our purpose. We
only require a good testing and training dataset for the program.
Steps for building Object Recognition Software:
1. Design the UI of the application.
2. Generating/Collecting Dataset.
3. Establishing the Neural Network and adding features
4. Train the System.
5. Test the System.
6. System testing i.e. testing all the modules together.

2.2. Architecture: -
When a user provides an image/object, then this image will be converted to a grey scale
image and then using Convolutional neural network (CNN) and Regional Proposal Network
(RPN) this image will get classified with the help of Spatial Pooling and a Classifier. The
resultant after all the above mentioned procedure will be that the application will recognise
the name or type of image/object the user will input and will showcase it on the screen.
3. Innovation:

The final application will able to recognise the particular types of objects and images
that the software have already been trained of. In future, it will able to recognise
more than 10 types of objects which is right now 3. Not only this, we will even
upgrade the software to recognise text images and phrases too by providing various
options in the software itself.
And this software can really boost the productivity of the local as well as MNC’s in
our country. We are also thinking to build a system that can recognise and convert
graphical data into text data or visa-versa.

4. Implementation: -

The overall implementation that has been done of the project is around 60%.
The name of the Object Recognition System is “Classy-Fi”. This system till now is
able to-
• get the user input image
• generate the data set
• convert image into a grey-scale image
• classify image using spatial pooling, regression, etc.
The front-end of the project is pending. The back-end tasks has been completed. The
project is completely working with UI.
4.1. Data-set Generation:
The data set has been taken from kaggle.com which provides some of the best
data-sets available. The data-set comprises two sub categories i.e.
1. TRAINING DATA
2. TESTING DATA
The training data set comprises of more than 40000 images and the testing
data includes around 20000 images.

4.2. Back-end Work ( Images ) :

Example of Training & Testing Data-set


CODE IMPLEMENTATION
CONVOLUTIONAL NEURAL NETWORK IMPLEMENTATION

*************************

You might also like