DeMe-Net

A CNN architecture to Diagnosis of Dermatology Melanoma Skin Cancer using parallel convolution.

The DeMe Net architecture is achieves 98.33% accuracy on testing data and 99.85% on the traininge detaset.

We used the HAM10000 dataset from Kaggle, which was published by ISIC for the 2018 ML Challenge.

The DeMe net architecture used 3x3, 5x5, 7x7, and 11x11 parallel conv2d layar for distributed computing.

We have used accuracy, precision, recall and f1-score as performance metrics to test the performance of the model.