Skin Diseases Classification Using Machine Learning

This was JSTP #22 project about using machine learning to classify most 10 skin diseases in Thailand but I get only 4 skin diseases in PJ6103_DATA

PJ6103_DATA

Non-dangerous skin diseases These classes no need to consult a doctor.

• Atopic Dermatitis [ad]
• Normal [nm]

Dangerous skin diseases These classes need to consult a doctor.

• Psoriasis [ps]
• Seborrhoeic Keratosis [sk]

Process

Train CNN model by using HAM10000 and transfer weights to PJ6103_DATA then ensemble the best 3 models for the best result.

About work

• For JSTP work
• For Best work

Usage

• Clone this respository

• Terminal: git clone https://github.com/filmerxyz/JSTP22_SkinDiseaseClassificationUsingMachineLearning

• Jupyter Notebook: !https://github.com/filmerxyz/JSTP22_SkinDiseaseClassificationUsingMachineLearning

Results

Confusion Matrix and Classification Report

After Ensemble using Arithmetic Mean with 3 models

Grad-CAM

Caution

Some line of code are wrong because I changed some filenames and I did not update those filenames, so if you want to run code you need to update some line of code.