Comparative evaluation of deep learning methods for masked face recognition
The thesis presents methods for improving the efficiency and accuracy of masked face recognition systems, which have become increasingly important in the aftermath of the COVID-19 pandemicThree different methods are proposed and evaluated, namely CNNSVM, PRFCNN, and HRNN. The CNNSVM method combines...
Saved in:
| 主要作者: | |
|---|---|
| 格式: | Thesis |
| 出版: |
2023
|
| 主題: | |
| 標簽: |
添加標簽
沒有標簽, 成為第一個標記此記錄!
|
| 總結: | The thesis presents methods for improving the efficiency and accuracy of masked face recognition systems, which have become increasingly important in the aftermath of the COVID-19 pandemicThree different methods are proposed and evaluated, namely CNNSVM, PRFCNN, and HRNN. The CNNSVM method combines CNNs and SVMs to extract features and classify masked face images, whereas the PRFCNN method enhances this process by incorporating a random forest, PCA and CNN. The HRNN method utilizes a recurrent neural network to improve the recognition of masked faces.. The proposed methods are evaluated and compared against several state-of-the-art approaches using the RMFD and LFW-SMFD benchmark datasets. The results show that all three methods outperform the existing methods in terms of accuracy, especially when high mask coverage is used.. The CNNSVM method achieved an accuracy of 98.39% on the RMFD dataset, outperforming the state-of-the-art methods. The PRFCNN method achieved an accuracy of 99% on the benchmark dataset, indicating the effectiveness of incorporating a random forest in the feature extraction process. The HRNN method, which leverages the temporal information in the masked face images, achieved a TAR of 99% on both benchmark datasets, demonstrating the potential of recurrent neural networks for masked face recognition. |
|---|