| Summary: | Skin cancer is one of the most common and fatal types of cancer that has shown a significant rise in patient numbers in recent decades. Skin cancer can be divided into both benign and malignant mutations. Early detection can help facilitate effective treatment before the skin cancer worsens. Therefore, a computer-aided diagnosis (CAD) system is essential in detecting skin cancer mutations. This study demonstrated a custom convolutional neural network (CNN) architecture and perspective transformation as image augmentation. The study explored the impact of image size and augmentation using perspective transformation on CNN performance. Experimental results revealed that image size (64×64) had the highest accuracy of 82.51% compared to others. Using the (64×64) image size, the perspective transformation as augmentation had the highest accuracy of 83.91% when the k-range is 8 and the number of augmented images is 20, showing the significance of large training sample sizes. This study confirmed the best image size to use and the impact of perspective transformation as an augmentation technique to enhance the CNN model performance. © 2024 IEEE.
|
|---|