Psoriasis detection and classification using MobileNetV3 and Coordinate attention

Abstract Psoriasis is a rare, persistent dermatological condition that requires careful monitoring to manage symptoms effectively. Its physical appearance increases the psychological trauma to those suffered with the disease. This research study primarily exploits the Deep learning (DL)-based model that leverages the MobileNet V3 network with coordinate attention (CA). A comparative analysis is conducted using pre-trained models, including VGG19, Ensemble RESNET, and MobileNet-V3. The models are tested using both the cross-entropy loss (CL) and the hybrid loss (HL). The incorporation of HL into the Adam Optimizer's objective function improved the model's accuracy. Notably, the accuracy increased from 93.20% to 98.40%, for the Mobilenet-V3 with CA and HL, the model can provide a reliable support in the detection and classification of psoriasis.