Deep Learning-Based Monkeypox Detection: A Hybrid Approach Using DenseNet121 and MobileNetV2

Abstract

Due to recent outbreaks outside of endemic areas, monkeypox is a newly developing zoonotic disease that has drawn international attention. As a result, early and precise lesion detection is crucial for effective containment and treatment. To overcome this difficulty, we suggest a hybrid deep-learning method that combines DenseNet121 and MobileNetV2 for reliable monkeypox identification from skin lesion photos using the Monkeypox Skin Lesion Dataset (MSLD) and the Monkeypox Skin Lesion Dataset V2.0 (MSLDV2.0). The accuracy of the proposed model was 98% on the MSLD2.0 dataset and 99.18% on the MSLD dataset. The model's robustness was further validated by the confusion matrix, which outperformed current techniques and demonstrated exceptional precision for both the monkeypox and non-monkeypox classes. The hybrid model makes it clinically feasible for environments with restricted resources by carefully balancing the computational efficiency of MobileNetV2 with the hierarchical feature extraction of DenseNet121. By providing a high-accuracy, deployable solution that supports international efforts to stop the spread of monkeypox through dependable early identification, this work advances AI-driven diagnostics.