Recognition of Type of Skin Disease Using CNN

According to a study conducted by National Centre for Biotechnology Information (NIH), the cost associated with lost productivity and treatment among those who sought medical care for skin cancer exceeded 1.2 billion dollars and also more than 5.1 million people got serious effects like infections, hair loss, itches, burns of skin cancer. The study also concludes that most of those cases can be decremented by early detection of the cancer. The diseases like basal cell carcinoma, melanoma, pyogenic granulomas, are cancerous diseases and non-cancerous diseases like dermatofibroma, melanocytic nevi, have a variety of harmful impacts on the skin and continue to spread overtime, if treatment of skin disease at early stage is not done then it leads to complication in the body and including spreading of the infection from one another. To overcome this an early detection of skin disease plays a very major impact in today’s world. Now a days image processing has become widely used in developing a solution to this type of problems. Developing a high accurate methodology can be used to decrement the count of skin infections and their huge loses. This paper presents a seven types of skin disease detection using CNN. The dataset used is HAM10000.We obtain high accuracy by making the dataset ordered by adding duplication. The input image undergoes different layers such as maxpool2d, conv2d, batch normalization, flatten, dense, Softmax etc.., As this classification is among seven different types of skin diseases out of which four are cancerous and other three are non-cancerous, the output is one among these seven.