Week 1: Skin Cancer Classification
Hi guys! We would like to give you brief information about the Machine Learning project in this post.
Our goal is to classify skin cancer using CNN on spiloma and mole.
To introduce ourselves, we are the DeepGuys: Mahmut Fatih Erbağ, Seda Oran and Furkan Gürel.
What is Skin Cancer?
Skin cancer is the uncontrolled growth of abnormal cells in the epidermis, the outermost layer of the skin, caused by unrepaired DNA damage that causes mutations. These mutations cause the skin cells to multiply rapidly and form malignant tumors.
Skin cancer is one of the most common cancers, the most dangerous type of cancer and very good results can be obtained when caught without spreading. If the disease is caught before the spreading cells pass through the bloodstream from one part of the body to another. 99% of people can get rid of the disease. However, if the disease spreads to other parts of the body, the rate of disease recovery decreases to 25%.When considering all these events we can use artificial intellligent on this subject.
Dataset
The main idea of project is classifiying skin cancer on people. There are two classes as benign and malignant. We are planning to combine the data we collect from ISIC-ACHIVE.This project will use artificial intelligence to classify mole and spiloma.Our goals to select this topis for increase skinc cancer deads with early diagnosis.When this project integrated in mobile phones people can use and know they are should go doctor or not. There are 21678 images in our dataset.
Solution Process
First of all, we will use CNN and the PyTorch framework.
We aim to achieve the most successful results by trying multiple pre-trained models. Of course, there will be changes in our methods against the problems that will arise during this solution. To give an example of these problems, the problem of overfitting may be to allocate the right number of train, validation and test images. Also, one of our biggest problems is that the dataset labels do not contain the same amount of images, so there is an imbalance in our dataset.
Have you seen these ?
Mahmut Fatih Erbağ — — — Furkan Gürel — — — VGG16(CNN)
— — — — — — — — — — — — Seda Oran — — — — — — — — —
