Achieving High Accuracy in Face Recognition: DeepFace Intervention and LBPH Comparison

Aim: The aim of this study is to achieve high accuracy in face recognition by comparing the performance of DeepFace with Local Binary Patterns Histogram (LBPH). DeepFace is proposed as an intervention to overcome the low accuracy issues associated with the LBPH algorithm. Materials and Methods: In this research, there are two groups. Group 1 uses a sample size of 26 for the LBPH algorithm, and Group 2 uses a sample size of 26 for the DeepFace model. The G Power value is 80% with a 95% confidence interval, and the threshold is 0.05%. The performance of both models is assessed using metrics related to accuracy, processing speed, and memory usage. The DeepFace model outperformed the LBPH algorithm, based on the results. Result: The DeepFace model outperformed the LBPH algorithm in terms of accuracy. DeepFace's accuracy range was 92.3% to 98.7%, whereas LBPH's range was 65.4% to 85.9%. Additionally, DeepFace outperformed LBPH in terms of processing speed and memory efficiency. The highest accuracy was recorded at a 95% confidence level and a significance level of 0.0042. Conclusion: The results show that for face recognition tasks, DeepFace outperforms the LBPH algorithm in terms of accuracy and efficiency. This study comes to the conclusion that using DeepFace can improve face recognition systems' efficiency and accuracy.