Skin issue is extremely normal in the day by day lives of people. Consistently a huge number of American individuals are influenced by skin issue of different types. Skin condition conclusion regularly includes a high level of information because of the scope of visual perspectives thereof. Since human judgment is constantly discretionary and infrequently reproducible, a PC helped indicative gadget ought to be considered for accomplishing an increasingly objective and precise conclusion. In this paper, we investigate the plausibility of utilizing profound Convolutional neural system (CNN) to make a widespread structure for determination of skin infection. We train the CNN engineering utilizing the Dermnet dataset's skin illness pictures and check its yield with both Dermnet and OLE, another information assortment for skin ailment, pictures. Our program can accomplish Top-1 exactness of up to 73.1 percent and Top-5 precision of 91.0 percent while running on the Dermnet dataset. Top-1 and Top-5 correctness’s for the OLE dataset check are 31.1 percent and 69.5 separately. We show that if all the more preparing pictures are utilized, those correctness’s can be additionally improved.