**4. Conclusion**

The deployment of technologies such as COVID-19 screening and counterfeit mask detection plays a crucial role in curbing the spread of the COVID-19 virus. This chapter has introduced advanced AI-driven methods designed to enhance both COVID-19 screening and the identification of face masks. Specifically, the utilization of deep learning for COVID-19 classification has demonstrated significant potential in the creation of effective screening tools. Moreover, the application of deep learning for face mask detection on mobile devices has exhibited promising results. In the pursuit of developing these innovative techniques, valuable insights are gained that can be applied in future pandemic situations. In the future, potential work includes: (1) Overfitting is characterized by machine learning models achieving impressive performance during the training phase, but faltering when tested on new data. To ensure the efficacy of large models for COVID screening and face mask detection, it is imperative to address this concern. In our forthcoming work, we aim to tackle this issue through the implementation of data augmentation techniques, leveraging expansive pre-trained image models. More specifically, these large pre-trained image models hold the capacity to generate synthetic data for both COVID screening and face mask detection. This synthetic data will encompass varying backgrounds, enriching the diversity of the training dataset and reducing overfitting tendencies. Additionally, we intend to incorporate human feedback as an integral part of the loop. This iterative process will involve human assessment to gauge the quality of the generated data, fostering continuous improvements in the data augmentation strategy. This combined approach of data augmentation and human feedback holds significant promise in enhancing the generalization capabilities of the models, thereby enabling robust and high-performing COVID screening and face mask detection systems in real-world scenarios; (2) AI ethics continue to be a significant concern across various applications, encompassing issues such as privacy breaches and biased data. In our future efforts, we are dedicated to addressing these ethical challenges head-on. Our strategy involves the application of privacy-preserving techniques to safeguard user privacy, coupled with measures to mitigate biased data. To protect user privacy, we are poised to employ our pioneering privacy-preserving edge intelligent computing framework. This entails training autoencoders in an unsupervised manner on individual edge devices. Subsequently, the latent vectors derived from these autoencoders are transmitted to the edge server for classifier training. This approach effectively reduces communication overhead while safeguarding end-users'sensitive data from exposure. In tackling biased data concerns, our plan is to integrate fair pre-processing techniques from AIF360, an AI fairness toolkit.3 These techniques will be strategically applied during the data collection phase

<sup>3</sup> https://github.com/Trusted-AI/AIF360

*Effective Screening and Face Mask Detection for COVID Spread Mitigation Using Deep… DOI: http://dx.doi.org/10.5772/intechopen.113176*

for both COVID screening and face mask detection. By doing so, we aspire to counteract biases that may emerge in the data, ensuring equitable and unbiased outcomes. By proactively addressing privacy and bias concerns through cutting-edge privacypreserving frameworks and fairness techniques, we aim to develop AI solutions that not only excel in performance but also uphold the highest standards of ethical conduct.
