AI Skin Cancer Detection GitHub

*DISCLAIMER: By downloading this tool, you understand that this is for educational purposes only. If you have a worry, please see a doctor. I am not responsible for any false positives or false negatives that this AI may predict.*

AI model to detect melanoma skin cancer

his a project made by me which utilises Artifical Intelligence (AI) to detect melanoma. This works by training a model on a huge dataset with benign and malignant images of suspected melanoma. The AI can then distinguish between the characteristics of a malignant and benign image. A suspected melanoma image can then be uploaded and ran against the trained model.

Features

  • Predicting melanoma by uploading a suspected image.
  • Upcoming: Re-training the model.

Dataset used

This model was built using the dataset Melanoma Skin Cancer Dataset of 10000 Images. This is a huge dataset consisting of thousands of images containing both benign and malignant images of melanoma.

Model Architecture

It uses a Convolutional Neural Network (CNN) to classify images of skin lesions as either benign or malignant.

How to use

  • Clone the repo here.
  • Unzip the model and put it in the same directory as the python file.
  • Run pip install -r requirements.txt to install the necessary libraries.
  • Run the main.py file
  • Upload a clear picture you want to test and get your result!

False Positives/Negatives

Whilst my model presents a high accuracy rate, there is still a chance that it may incorrectly classify an upload. As per the disclaimer at the top, please do not take the result of this model with any confidence, as it is not approved by any 'medical institution'. To reiterate, if you have an issue, PLEASE SEE A DOCTOR.

Training model option - not ready yet

There should not be a need to re-train the model as there is already a trained model in the directory. It is also not recommended to train the model again because it takes so long. But if you have to, just click the 're-train model' button and wait until it is done.

Utilising this model in medical technology

Whilst this model is not currently fit for accurate predictions in the real-world, it still presents a high accuracy rate. This model with some adapting, is an example of what can be achieved with artifical intelligence to help patients get an answer quicker.