Skin disorders pose a grave threat to global public health. The increasing prevalence of skin disorders has placed a strain on the public health system. Due to the scarcity of specialists, it is necessary to develop automated methods that can quickly detect the disease in order to alleviate patients' financial and health concerns. Using approaches based on deep learning, it is feasible to diagnose various potentially serious skin disorders in their earliest stages. This study explores the fundamentals of deep learning model development for dermatological classification issues and provides recommendations. In addition, the scarcity of medical data pertaining to darker skin tones and demographic imbalances in dermatology are briefly addressed. Even-though deep learning frameworks are growing, but there are many unexplored obstacles and possibilities in this sector.