<p>Skin diseases commonly affect different age groups and with different groups of conditions. The severity of skin disease varies according to diverse conditions, with normal concerns such as acne and eczema, and serious diseases like cancer. However, there is a need for large and standardized datasets, image variations due to lighting effects, and high similarities among the diseases have a major impact on overall efficiency in traditional techniques. Thus, an advanced segmentation and classification model is necessary to achieve highly effective dermatological care. Therefore, this work designed an innovative deep learning-enabled skin disease classification model to alleviate the issues in dermatological image analysis. The skin images from publicly available sources like HAM10000 and PH2 Dataset are collected. These collected images are further fed into the developed Full Resolution Residual Networks with Attention Mechanism (FRRNet-AM) to segment abnormal regions from the image. The developed FRRNet-AM identifies accurate details of lesion boundaries, and it captures the subtle variations in the image. Then, various skin disease types are classified using the developed Adaptive Mobilenet (AMNet) framework, where several parameters are optimized by the novel Modified Random Function-based Secretary Bird Optimization Algorithm (MRF-SBOA). The developed AMNet classification module finds different skin disease classes with a high accuracy of 97.33% and a precision of 96% using dataset 1. Experimental validation is conducted against standard works, and the results prove the model’s efficiency in dermatological image analysis.</p>