/challenge-mole

Primary LanguagePureBasic

SKIN CANCER DETECTION USING CNN AND KERAS PROJECT

Description

Skin Cancer Definition and Types:

  • Most skin cancers are locally destructive cancerous (malignant) growth of the skin. They originate from the cells of the epidermis, the superficial layer of the skin and the majority of them rarely spread to other parts of the body and become life-threatening, ecxept for Melanoma which is lethal.

  • There are three major types of skin cancer: (1) basal cell carcinoma (the most common), (2) squamous cell carcinoma (the second most common) and (3) melanoma

  • The images used for this CNN excercise cover two cancerous types; bcc and mel,together with another 5 benign types; Akiec,Bkl,Df,Nv,and Vasc

    Abbreviation Full name
    Bcc Basal cell carcinoma
    Mel Melanoma
    Akiec Actinic keratoses and intraepithelial carcinoma
    Bkl Benign lesions of the keratosis
    Df Dermatofibroma
    Nv Melanocytic nevi
    Vasc Vascular lesions

Project mission:

  • The main mission of this project was to create an app using CNN and Keras to diagnose the mole images. The app could predict whether the image detected/loaded is cancerous or benign

NocCan App

Website

https://melanomia-detection.herokuapp.com/ https://share.streamlit.io/minhhienvo368/cancer-detector/main/mole_detection.py

Visuals:

Snapshot of the App

Installation on local machine

*On Windows $ virtualenv venv $ \venv\scripts\activate

Or if using Linux/ MACOS $ python3 -m venv myvenv $ source myvenv/bin/activate

Install the requirements: $ pip install -r requirements.txt

Run the app: $ python app.py

Python version

  • Python 3.9

Packages used

  • os
  • numpy==1.19.5
  • pandas==1.3.3
  • matplotlib.pyplot==3.4.2
  • itertools
  • seaborn
  • sklearn
  • tensorflow.keras
  • Keras==2.4.3
  • Pillow==8.3.1
  • scikit-learn==0.24.2
  • streamlit==0.88.0
  • tensorflow-cpu==2.5.0

Usage and links

File Description
mole_detection.py Main python code
mole_model.py Python code with Neural network model
mole_preprocessing.py Python code for visuals (dataset and results)
visuals Folder including the plots presented on the Readme
NoCan App Presentation Powerpoint presentation: https://docs.google.com/presentation/d/1IPsoEJlo0RAcnG8BPl25Rd-0FjspsNQpU9mjD7WalmM/edit#slide=id.g1f87997393_0_864

The Dataset

The dataset used for the model can be found at https://www.kaggle.com/kmader/skin-cancer-mnist-ham10000 It was created by Tschandl et al. 2018.

Contributors

Name Github
Amaury van Kesteren https://github.com/AmauryvanKeste
Heba Elabrak https://github.com/Helabrak
Michel OMBESSA https://github.com/mdifils
Minh Hien Vo https://github.com/minhhienvo368

Timeline

20-09-2021 to 24-09-2021