Modifikasi Convolutional Neural Network Arsitekter VGG16 Dengan Dull Razor Filtering Untuk Klasifikasi Kanker Kulit

Skin is an important element in the human body which plays a role in protecting internal organs from direct exposure to sunlight or ultraviolet light. Sunlight is a source of vitamins for humans, but excessive exposure can damage skin cells and potentially cause skin cancer. This disease develops in the upper layers of the skin and can be recognized by changes in the skin, such as lumps or abnormal growths. Skin cancer is a significant health problem, and the use of CNNs has proven effective in diagnosing this disease through image analysis. However, this study attempts to improve the classification accuracy by introducing a modification to VGG16, which involves the use of the Dull Razor Filtering technique. Dull Razor Filtering is an image processing method that aims to reduce noise and improve feature extraction in images. Its application to the VGG16 architecture aims to optimize the identification of more complex skin cancer patterns. Experiments were conducted using a skin cancer dataset that included a variety of lesions and skin cancer types. Therefore, the Deep Learning method, especially using VGG 16 with Dull Razor Filtering, has succeeded in increasing the accuracy level of skin cancer classification. In experiments with 10615 images, the results show the highest level of accuracy reaching 92.88% after going through the training process with the VGG 16 model.