DeepAc-Web: A Web Based Intelligent System for Acne Severity Grading

Acne vulgaris is the most common skin disorder that involves blockage and inflammation of hair follicles and dead skin cells resulting from increased sebum production. It affects people of all ages, skin types, and backgrounds, with a particularly significant impact on adolescents worldwide. Precise grading of acne is crucial for determining effective treatment. However, assessing acne severity is challenging due to the similar appearance of all kinds of lesions and the difficulty of reliably counting them. Therefore, developing an automated system has become quite essential for fast, accurate, and timely diagnosis of acne. In this paper, we propose to develop a deep learning-based model that can correctly grade acne severity, as well as to implement an intelligent web-based system to predict the class of acne image. Initially, we designed a ResNet-152 model integrated with the Self Attention mechanism, utilizing the ACNE04 dataset containing 1457 clinical images. After evaluating with the dataset, the proposed model outperforms all state-of-the-art methods, with a significant accuracy of 77.05%. Finally, we have developed a web-based application (DeepAc-Web) by deploying the proposed model using Streamlit API. This system is suitable to use in clinical settings since it is time-efficient and reduces inter- and intra-observer inconsistencies. With the help of this remotely accessible technology, early acne diagnosis can be achieved for better treatment and patient outcomes.