You are on page 1of 14

SURESH ANGADI EDUCATION FOUNDATION’S

ANGADI INSTITUTE OF TECHNOLOGY AND MANAGEMENT


Savagaon Road, BELAGAVI – 590 009.
(Approved by AICTE, New Delhi &Affiliated to Visvesvaraya Technological University, Belagavi, Accredited by NAAC)

DEPARTMENT OF ARITICIAL INTTEELLIGENCE AND DATASCIENCE

PROJECT PHASE-2

ENHANCEMENT OF RAILWAY SAFETY MEASURES THROUGH DEEP LEARNING


ALGORITHM TO IDENTIFY RAILWAY WHEEL DEFECTS

SUBMITTED BY- PROJECT GUIDE-


Mr. HATE NAGARAJ GOUDA
Mr. SHASHANK D. VIGHNESHI PROF. Chetan PATIL
Ms. MEGHANA CHITTAPUR
Mr. Tejas Nandani
INTRODUCTION :
• The project is railway wheel defect detection using deep learning algorithms for
finding the defects on the railway wheel surface.
• The model allow user to upload images of defected wheel by image processing it
detect the wheel is defected or not.
• Through the integration of deep learning algorithms, we improve the efficiency
of detection system for enhance safety measures.
• Automating the defect detection process, our project streamlines maintenance
procedures and reduces reliance on manual inspection methods.
PROBLEM STATEMENT :
The railway industry is confronted with the challenge of ensuring the safety and reliability of railway
operations by effectively detecting defects on railway wheels. Traditional manual inspection methods
are time-consuming, labor-intensive, and prone to human error, leading to potential safety hazards and
operational disruptions. Therefore, there is a pressing need for a robust and automated solution that can
accurately detect defects on railway wheels
OBJECTIVES :
 Integrate a deep learning algorithm, such as Convolutional Neural Networks (CNNs), to
process the captured images, train the algorithm to accurately identify and classify different
areas of defects on railway wheels, such as flat spots or cracks.
 Enable real-time processing of images to provide immediate feedback on the presence and
severity of defects, allowing for timely actions.
 Achieve a high level of accuracy and reliability in defect classification, ensuring that the
algorithm correctly identifies and prioritizes major areas needing attention.

.
SYSTEM REQUIREMENTS:
Hardware Requirements : Software Requirements :

 Data Preprocessing Tool : OpenCV, Numpy, Pandas


 Processor Core : i5 core  Database : MongoDB, SQLite

 Operating System : Windows 10 or above / Linux Based


 Processor speed : 2.11 GHz (Ubuntu)

 Coding Language : Python(3.11.8)


 Development : Jupiter Notebook, VS Code
Environment

 Hard disk : 256GB  Deep Learning : Tensorflow, PyTorch, Keras


Frameworks
ARCHITECTURE DIAGRAM:
TRAINING
USER
AND TESTING

UPLOAD CONVOLUTIONL
DATA PARAMETER
RAILWAY WHEEL NEURAL
PREPROCESSING TUNING
IMAGES NETWORK

ANDROID GOOGLE
API KEY CLOUD PICKLE
APPLICATION

IMAGE
SHOWS
PROCESSING
USE CASE DIAGRAM:
LOGIN

UPLOAD IMAGE

TRAINING MODEL

IMAGE GOOGLE
PROCESSING
CLOUD
USER
CLASSIFICATION

DAMAGE
ANALYSIS

API KEY

REPORT
BLOCK DIAGRAM:

DATA DATA DEEP


COLLECTION PROCESSING LERNING

FEATURE
OUT PUT CLASSIFICATION
EXTRACTION
ACTIVITY DIAGRAM:

LOGIN

UPLOAD GENERATE INPUT LIVE VIEW


CLASSIFICATION
DATA REPORT DATA REPORT

MODIFY
UPLOADED

LOGOUT
SEQUENCE DIAGRAM:
:Data
:Data Input :Model Selection :Training and Testing :Parameter Tuning Pickle
Preprocessing

Removing
Choosing Train the model serializing and de-
Data Uploading unwanted data
algorithm and and assigning the serializing a Python
and noise
Data splitting parameters object structure
DATAFLOW DIAGRAM:
START

UPLOAD IMAGE
IMAGES PROCESSING

MODEL
RESIZE THE TRAINING
IMAGE AND TESTING

INITIALIZE
FEATURE
CONVUTIONAL
EXTRACTION
NEURAL NETWORK

GOOGLE
API KEY CLASSIFICATION
CLOUD

APLICATION REPORT
CONVOLU Assigning
DATA Cleaned data TIONL parameters PARAMETER
PREPROCESSING NEURAL TUNING
NETWORK

model
Selection of
Trained
Model
TRAINING
AND TESTING

Fig: DFD Level 0


GOOGLE
PICKLE API KEY
CLOUD

IMAGE
PROCESSING
Thank You!

You might also like