Scribd
Upload
6 views

Main DRC (1) - 1

The document proposes a deep learning system using ResNet-50 and EfficientNetB0 algorithms to detect cataract disease from retina images. It discusses the need for an automated solution, literature review, existing methods, proposed architecture, data flow, models, and concludes that deep learning can provide accurate and efficient cataract detection.

Uploaded by

kirankohli040600
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
6 views

Main DRC (1) - 1

The document proposes a deep learning system using ResNet-50 and EfficientNetB0 algorithms to detect cataract disease from retina images. It discusses the need for an automated solution, literature review, existing methods, proposed architecture, data flow, models, and concludes that deep learning can provide accurate and efficient cataract detection.

Uploaded by

kirankohli040600
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 23

CATARACT DISEASE DETECTION USING DEEP

LEARNING
PROJECT DRC-1 PRESENTATION

CATARACT DISEASE DETECTION USING DEEP LEARNING


By
POTUPUREDDY SURYA KIRAN
Reg. No. 322203320048
IV Semester. M.C.A
Under the guidance of
Mr. P. V. V. R Chandra Sekhar
Assistant Professor
Department of Computer Applications

DEPARTMENT OF COMPUTER APPLICATION


GAYATRI VIDYA PARISHAD COLLEGE OF ENGINEERING
(AUTONOMOUS)
(Affiliated to Andhra University, A.P)
VISAKHAPATNAM – 530 048
2022-2024
ABSTRACT

As we all know eye is an important sensing organ of human being. Eye is interconnected with
parts like iris, pupil, lens, retina, cornea, optic nerve. Cataract is one of the most common eye
diseases which may leads to blindness mostly the main symptom is blurry vision. Cataract is an
age-related eye disease, and its incidence rate rises with age, but it occurs to new born babies in
rare cases. Early intervention and timely treatment can largely avoid blindness. The proposed
model is going to use the deep learning algorithms such as ResNet-50 algorithm and
EfficientnetB0 algorithm. The proposed system will detect the cataract and helpful to the people .
PROBLEM STATEMENT

Cataract disease, if not detected early, can lead to severe vision impairment. The current methods
of detection may involve time-consuming manual processes and are subject to interpretation
variations. This project addresses the need for an automated system that utilizes deep learning
techniques for efficient and accurate cataract disease detection.
LITERATURE SURVEY

[1] International Journal of Research Publication and Reviews, Vol 4, no 4, pp 2130-2136 April 2023
2131
Isma Shaheen - To alleviate the burden of ophthalmologist, many researchers working in the field of
biomedical imaging developed a number of techniques for the automatic detection and grading of
cataract. Imaging modalities used for this purpose includes slit-lamp images, retro-illumination
images, digital/optical eye images, retinal images, and ultrasonic Nakagami images.
[2] Linglin Zhang - To investigate the performance and efficiency by using Depp Convolutional
Neural Network (DCNN) to detect and grad cataract automatically, it also visualize some of the
feature maps at pool5 layer with their empirical semantic meaning, providing a explanation to the
feature representation extracted by DCNN.
Existing System

Current methodologies for cataract detection often rely on manual interpretation of eye images or
traditional computer vision approaches. While automated systems exist, they may lack the accuracy and
robustness required for precise cataract identification. There is a need for an advanced system that
leverages the deep learning algorithms to enhance the efficiency and accuracy of cataract disease
detection.
PROPOSED SYSTEM

The proposed system introduces a sophisticated approach by integrating ResNet-50 and


EfficientnetB0 algorithms for cataract disease detection. Python, along with Streamlit, will be utilized
to create a user-friendly interface. This advanced system aims to provide accurate and efficient
cataract diagnoses, enabling timely medical interventions.
SYSTEM REQUIREMENTS

Software requirements:
Front End : Streamlit
Python(3.8)
Google Colab
Visual Studio Code
ARCHITECTURE

Data Collection

Preprocessing

lTraining Testing

ResNet-50 and EfficientnetB0


Models

Input Retina Image

Model Evaluation

Cataract Detection

Display Results
DATA FLOW DIAGRAM
Level 0

Input Retina Cataract


User Output
Image Detection
USE CASE DIAGRAMS
RESNET-50 MODEL :

 ResNet-50 tackles gradient vanishing, critical for deep learning, via skip connections.

 Residual networks use the concept of skip connections. Using these skip connections, we can resolve the problem

of vanishing gradient.

 Skip connection means adding original input to the output by skipping some convolutional layers.

 ResNet-50's blocks sum input with convolutional block output, facilitating learning from input-output disparities.

 ResNet-50 transitions learning from output (Y) to minimizing difference (F(x)), aiming for Y ≈ X.

 With 50 layers structured into blocks, ResNet-50 optimizes feature extraction and learning for cataract detection.
EFFICIENTNETB0

 Compound scaling method scales depth, width, and resolution simultaneously.


 Architecture combines depthwise separable convolutions, batch normalization, and Swish
activation.
 Depthwise separable convolutions reduce parameters and computations
 Network designed for maximizing feature reuse and minimizing redundancy.
 Training utilizes stochastic gradient descent (SGD) with momentum and weight decay.
 Data augmentation techniques like random cropping and color distortion are applied.
 Trained on large-scale datasets such as ImageNet for image classification tasks.
 Fine-tuning and transfer learning allow adaptation to specific tasks or datasets.
 EfficientNetB0 achieves state-of-the-art accuracy with lower computational cost.
SAMPLE CODE
OUTPUT:
CONLUSION

In conclusion, deep learning for cataract detection offers high accuracy, automation, and early intervention
potential. Challenges like dataset bias and interpretability remain. Collaborative efforts are crucial for clinical
validation and regulatory approval, promising improved diagnosis efficiency, accessibility, and patient outcomes
in the future.
THANK YOU

You might also like