Table 6.
Comparative analysis of various SOTA models/techniques used for face mask identification FMD Task.
| Author | Year | Model | Performance |
|---|---|---|---|
| [21] | 2020 | Deep learning framework | Accuracy: 79.24% |
| [22] | 2020 | Deep learning framework | Accuracy: 98.70% |
| [23] | 2020 | Deep learning and classic projective geometry techniques | AUC: 97.6%, precision: 97.00%, recall: 97.00% |
| [24] | 2020 | Deep-Learning-based SRCNet | Accuracy: 98.70% |
| [26] | 2020 | HGL to deal with the head pose classification with CNN | Front accuracy: 93.64%, side accuracy: 87.17% |
| [29] | 2020 | Deep learning method called FaceMaskNet | Accuracy: 98.6% |
| [30] | 2020 | Generative adversarial networks (GANs) and support vector machines (SVMs) classifier | Mask sub-challenge: 74.6% |
| [31] | 2021 | ResNet50, AlexNet, and MobileNet | Accuracy: 98.2% |
| [32] | 2022 | Expanded mask R-CNN | mAP: 80.25 |
| [34] | 2022 | Deep MaskNet framework-MDMFR | Accuracy: 93.33 |
| [35] | 2022 | CNN architecture | Accuracy: 98% |
| Proposed Model | FMDNet | Accuracy: 99% | |