https://github.com/biswa932/eczemadetectionusingcnn

Machine Learning Model for Skin Condition Detection using CNN
https://github.com/biswa932/eczemadetectionusingcnn

cnn deep-learning machine-learning tensorflow

Last synced: 3 months ago
JSON representation

Machine Learning Model for Skin Condition Detection using CNN

• Host: GitHub
• URL: https://github.com/biswa932/eczemadetectionusingcnn
• Owner: biswa932
• Created: 2023-12-28T14:40:04.000Z (over 2 years ago)
• Default Branch: main
• Last Pushed: 2024-01-08T20:27:07.000Z (over 2 years ago)
• Last Synced: 2025-08-26T00:35:04.506Z (10 months ago)
• Topics: cnn, deep-learning, machine-learning, tensorflow
• Language: Jupyter Notebook
• Homepage:
• Size: 62 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# Eczema Detection using 3 CNN model stages:

Three stage detection of eczema

This program uses 3 stages/ models:

1. Human Skin Detection model : if passed image is of human skin or not

2. Eczema detection model: if passed image of human skin has eczema or not

3. Eczema classification model: the level of eczema (mild, moderate, severe)

# Pre-trained model used: 

https://www.kaggle.com/models/google/mobilenet-v2/frameworks/TensorFlow2/variations/140-224-feature-vector/versions/2

This pre-trained model "mobilenet-v2" has been retrained using new dataset and a single dense layer for each case.

# Eczema images Dataset:

The eczema dataset was downloaded from: https://www.kaggle.com/datasets/shubhamgoel27/dermnet

# Dataset structure: dataset/eczema_photos

    1. allSkin: 2612 items

    2. clearSkin: 1204 items

    3. eczema: 1408 items

    4. invalid: 2400 items

    5. mild: 393

    6. moderate: 641

    7. severe: 373

# Accuracy:

1. Human Skin Detection model : Training(loss: 0.0086 - acc: 0.9995), Evaluation(loss: 0.0225 - acc: 0.9935)

2. Eczema detection model: Training(loss: 0.0690 - acc: 0.9752), Evaluation(loss: 0.0823 - acc: 0.9674)

3. Eczema classification model: Training(loss: 0.3789 - acc: 0.8927), Evaluation(loss: 0.8193 - acc: 0.6097)

# All the models were converted to CoreML (*.mlmodel) using Colab.

1. Code: ThreeStageEczemaDetection.ipynb

2. CoreML convertion: ThreeStageEczemaDetectionToCoreML.ipynb