Published Paper Details:

The uploaded document contains a project on developing a hybrid Vision Transformer (ViT)-CNN model for face mask detection. Here's a 200word abstract based on the details provided: This project introduces a novel hybrid Vision Transformer (ViT)-CNN model for efficient face mask detection, leveraging the strengths of convolutional neural networks (CNNs) and pre-trained Vision Transformer backbones. The approach involves local feature extraction using CNN layers combined with the global feature representation provided by the EfficientNetB0 model. The dataset, structured into masked and non-masked categories, is preprocessed with data augmentation techniques to enhance generalization. The hybrid model integrates CNN-based hierarchical feature maps with transformer-derived global features using concatenation. The architecture includes fully connected layers for binary classification and is trained with an Adam optimizer and binary cross-entropy loss function. Key metrics such as training and validation accuracy are monitored across epochs. The system demonstrates robust performance with a test accuracy of over 90%, validating its applicability in real-world scenarios. Further, the model supports individual image predictions, offering a userfriendly interface for detecting mask compliance. This project emphasizes the synergy between CNN and transformer models in addressing critical healthcare challenges. Future work could explore multi-class classification or real-time deployment for public safety monitoring.