Skin cancer comes in many different forms, including skin lesions. Melanoma is the primary focus of this paper among the various kinds of skin cancer. When this type of cancer is detected at an early stage, it is treatable. A computer-aided recognition is necessary for the earlier diagnosis. The standard image segmentation methods are used in a variety of computer-aided recognition techniques that are now accessible. In this research, we suggested a new image segmentation approach called the MR U-Net model. We will separate melanoma tumors out of the dermoscopic images using this technique. In the system that is being presented, deep learning is also used. This allows us to initially train the systems with just a handful of datasets and then test it with an extensive amount of datasets. The system's effectiveness is assessed based on the following factors, including detection accuracy and system accuracy.