Human Skin Disease Detection System Using CNN

Project Domain / Category

Software Application

Abstract / Introduction

Skin diseases are widespread, affecting millions of people globally. Many individuals suffer from various types of skin conditions that often pose hidden risks, leading not only to self-esteem issues and psychological distress but also increasing the risk of skin cancer. According to the World Health Organization (WHO), approximately 30% to 70% of the population is affected by skin diseases, with many lacking awareness about the different types.

To address this issue, you are required to develop a Skin Disease Detection System using Convolutional Neural Networks (CNN). The goal of this project is to provide users with an easy way to identify potential skin diseases, allowing them to take proactive measures for treatment. It also assists doctors by offering preliminary insights into the type of skin condition, which helps streamline and improve the efficiency of diagnosis.

The system will have two types of users.

1.      End User

2.      Admin

1.      End User

The End users must register and log into the system. Once logged in, they can upload an image of their skin condition, and the system will automatically classify the disease shown in the image. Additionally, users can view a list of doctors based on the identified skin condition and provide can provide feedback.

2. Admin

The admin has the ability to log in with their credentials, can manage users(add/update/delete), and can manage the doctors by adding, updating, or removing entries. They can also access the feedback submitted by users.

Functional Requirements:

The project will comprise of the following functional requirements: -

1.      Data pre-processing

Image data pre-processing is a crucial step that can significantly influence the performance of your model. Images of skin lesions are collected and annotated with disease classifications. To enhance the dataset, data augmentation techniques create variations of existing images. Next, images are standardized in size to ensure uniform input for the model and pixel values are normalized to a consistent range. The dataset is then divided into training, validation, and test sets. To address class imbalance, techniques may be employed to manage underrepresented diseases.

2.      Model training and testing

After pre-processing, the dataset is ready for training. You're required to distribute 80% data for model training, so the model can adapt to the maximum of cases, while 20% data for testing.

3.      Model tuning