The following research work is intended to enhance the efficiency of the facemask detection system, which is important in limiting airborne diseases transmission, especially in places where the rate of infection is most likely, such as hospitals wo approaches are proposed in this paper for enhancing surveillance: the first model is a custom system model using convolution neural network (CNN), which gave high sensitivity and 96.4% accuracy and The second approach is a hybrid model system that use CNNs for feature extraction along with a pre-trained classifier algorithm Darknet. This hybrid method leverages the strengths of both CNNs and pre-trained algorithms improved accuracy, stability, and reduced loss. These results clearly indicate that the best performance in terms of accuracy and stability is achieved using the hybrid model system by reaching accuracy 98% This model is sensitive to delay and thus highly adaptable across different datasets as it trained on a huge dataset with verity of images hence, we suggest using it on a 6G network at an estimated data rate of one terabit per second, and taking the advantage of visible light communication (VLC) especially in hospitals as its more safe for human's health. These will provide valuable inputs not only for research in the future but also hold immense promise for greatly improving practical applications in image classification tasks.