Covid prevention based on identification of incorrect position of face-mask.

Since the rise of the COVID-19 pandemic, safeguarding public health has taken paramount importance, drastically transforming lives in unprecedented ways. Mitigating the virus's spread necessitates various measures, including wearing face masks. However, there has been a concerning trend of laxity and incorrect use of masks in public spaces, heightening the fear of escalating cases globally. Numerous deep-learning models have been proposed for face mask detection. This paper presents s novel DenseMaskNet deep learning model which integrates real-time visualization with transfer learning techniques. This involves a two-step process. The first step is mask detection, which is built by fine-tuning the state-of-the-art pre-trained deep learning model, DenseNet201. The Face- Mask Recognition Dataset (FMRD), which is free from biases and contains masked, unmasked, incorrect masked faces, is utilized in the mask detection process. In the second step, Face detection and real-time visualization, prediction scores calculated for each image of the FMRD dataset are fed into a res10_300 × 300_ssditer_140000_caffemodel, a pre-trained single-shot MultiBox face detector model, which combines these scores to enable real-time face mask detection. The OpenCV library is utilized to capture real-time video from a camera and experiment with our model in practical scenarios. Comparative analysis against VGG16, InceptionV3, ResNet50, MobileNetV2, and Efficient-NetB7 reveals that DenseNet201 consistently outperforms all metrics. Notably, DenseMaskNet exhibits superior performance with precision, F1 score, recall, and accuracy of 98%, 98%, 99%, and 99%„ respectively. These compelling results affirm the efficacy of DenseMaskNet for real-time face mask detection, suggesting its viability for widespread utilization in real-world settings. [ABSTRACT FROM AUTHOR]