/SkinAI

FDAA_Team6_XingYun_MehtaNguyenVerma

Primary LanguageJupyter Notebook

SkinAI - SC1015 - FDAA_Team6_XingYun_MehtaNguyenVerma

Introduction

This is our Mini-Project for Introduction to Data Science and Artificial Intelligence (SC1005). Our dataset was sourced from Kaggle, which contains labeled images of various skin conditions.

Practical Motivation

Early and accurate diagnosis of skin conditions can be challenging due to the wide variety of conditions and their similar visual manifestations. Using a trained model can enable users to detect these conditions without having to get it checked constantly.

Problem Definition

Skin Disease Detection - Image classification of skin conditions, aiming to assist in automated detection based on visual symptoms

Objectives

  1. Develop a machine learning model using convolutional neural networks (CNNs), KNN, Random Forest Classifiers, and Naive Bayes to classify skin diseases.
  2. Achieve high accuracy in disease identification
  3. Explore different techniques for data preprocessing, evaluation (confusion matrix, accuracy score, etc), and deployment of the trained model.

Required Python libraries

  • NumPy
  • Pandas
  • Matplotlib
  • Seaborn
  • Scikit-learn
  • TensorFlow
  • Keras

Machine learning models used

  • Convolutional Neural Network
  • KNN
  • Random Forest Classifier
  • Naive Bayes

Skills learned from this project

  1. Handling inconsistent datasets for data cleaning and standardization of image size, format, and color channels. Resizing of images for memory efficiency and faster processing
  2. Identifying suitable classes for model training
  3. Development of various machine learning models for image classification tasks
  4. Understanding of the concepts of metrics used for model evaluation
  5. Collaborating effectively on GitHub and Google Colab with free GPU access
  6. Deploying a trained model using Streamlit
  7. Problem-solving and troubleshooting: Learning how to efficiently debug code, address various issues, and optimize the model's performance.

Contributors

  1. Nguyen Hoang Minh Ngoc - U2323871H @angelinawong1210
  2. Verma Shireen - U2323013D @s812v
  3. Mehta Rishika - U2323133H @Oganesson0221

Driven by the shared goal of learning and expanding our knowledge about each machine learning model, we collaborated closely, pooling our skills and perspectives to tackle challenges, explore new techniques, and collectively deepen our understanding of image classification.

How to try the model?

You can just clone this repo into your personal devices, installing all essential modules and run the script streamlit run 7-UI.py

References