Selection of Facial Geometric Key Points using Convolution Neural Network

Abstract

Face Recognition is a significant authentication procedure that is being used in bio-metrics in these days and age. Face recognition is based on various algorithmic approaches. Deep learning-based neural networks have a significant influence on the facial recognition approach. VGG (Visual Geometric Graphic) based networks are the multiple layered convNet is a 224*224 RGB channel based facial network built on various face samples. The major feature of this paper is the development of the facial key markers from facial features the procedure is conducted on samples of various data sets. The samples are then processed into the VGG19-based neural network for the facial detection of the samples processed from the data sets. Then the facial key features are extracted from the faces that are processed from the samples. The facial key features are based on the faces\' eyes, forehead, and lip regions. Then the facial key markers are identified on the various faces. Then the facial geometry is calculated based on the facial key markers that are located on the individual faces the calculated facial geometry is then applied to the facial recognition of the sample of the data sets. This facial geometry is the basis of deep facial recognition on the facial samples. The face recognition that is conducted has an accuracy of 98%. Facial recognition is the basis of various authentication procedures for web-based login. The developed methodology is a low-cost computational procedure to conduct accurate facial recognition on commodity hardware.

Introduction

The text focuses on the development and methodology of facial recognition systems using biometric and deep learning approaches, emphasizing accuracy, feature extraction, and geometric analysis for individual identification. Facial recognition is widely used in security, authentication, human-computer interaction, and surveillance, and its effectiveness relies on accurately detecting and analyzing unique facial key features.

Key Points:

1. Significance of Facial Recognition:

   • Facial recognition is a high-precision biometric system that identifies individuals based on unique facial structures.

   • Applications include home security, access control, and business authentication systems.

   • Advances in sensors and imaging technologies have enhanced the accuracy and reliability of face detection.

2. Feature Extraction Approaches:

   • Traditional methods rely on geometric measurements between facial points (e.g., eyes, mouth, nose) and linear algebra techniques like Eigenvalues, PCA, and Independent Component Analysis.

   • Neural network-based methods, particularly Convolutional Neural Networks (CNNs), can learn hierarchical features from images for better recognition accuracy.

   • Modern approaches use pretrained deep learning models, such as VGG-19, to extract meaningful facial features and reduce irrelevant components using Principal Component Analysis (PCA).

3. Datasets:

   • The system is tested on YALE and ORL datasets:

     • YALE: 800 images of 40 individuals, 20 samples per person.

     • ORL: 2,000 images of 200 individuals, 10 samples per person.

   • These datasets provide multiple angles and facial variations for robust feature learning.

4. Methodology:

   • Facial key regions include the forehead, eyebrows, nose, lips, and cheeks, which are unique for each person.

   • Images are preprocessed into 255×255 RGB channels, then passed through VGG-19 CNN layers for hierarchical feature extraction.

   • Micro-patterns of faces are processed in convolutional and max-pooling layers with progressively smaller resolutions (128×128 → 64×64 → 32×32) to capture fine-grained features.

   • Extracted patterns are reassembled through fully connected layers and Softmax layers for final recognition.

   • Optimization involves cluster analysis and Lagrange multipliers to maximize accuracy in identifying facial key features.

5. Algorithm Steps:

   • Input images are processed through VGG19 layers, subdividing into micro-resolutions and detecting facial patterns.

   • Max-pooling layers reduce dimensionality while preserving essential features.

   • Features are stored, reassembled, and processed through fully connected and Softmax layers for final identification.

   • The system effectively distinguishes individuals by comparing extracted facial key features across the dataset.

6. Related Work:

   • Fuzzy Neuro Inference Systems: Achieve 92% recognition accuracy.

   • Bayesian Multi-Face Recognition: Enables low-memory real-time detection.

   • PCA-based Systems: Fast and effective, but limited in varying backgrounds.

   • VGG16 CNN Pipelines: High accuracy (94.4%) even on small datasets, simpler than deeper CNNs.

Conclusion

The system developed which is discussed in this paper was found to be efficient as seen in the above table. The facial geometry calculation from the above method is an efficient way of deep facial recognition. The memory allocated for each node is very less compared to the previous methods. The system primary memory is also consumed very less compared to the other methodologies so this will lead to low computational cost consumption. The image matrix size is very low compared to the VGG16 so this leads to the adaptability of the system is more likely since it can be deployed with less computational cost. As discussed above the systems testing parameters which include accuracy increases significantly, by this analogy it can be said that the system can be implemented on commodity hardware which is of low cost so the authentication system developed on top of this methodology is very much economical. Facial geometry can be successfully applied to the bio-metric authentication process with more accuracy and less computational cost which makes this system more efficient.

References

[1] A. Bhatia, S. SriVastava and A. Agarwal., \"Face Detection Using Fuzzy Logic and Skin Color Segmentation in Images\",3rd International Conference on Emerging Trends in Engineering and Technology, 2010, pp. 225-228,doi: 10.1109/ICETET.2010.20. [2] E. B. Putranto, P. A. Situmorang and A. S. Girsang., \"Face recognition using Eigen face with naive Bayes\", 11th International Conference on Knowledge, Information and Creativity Support Systems, PP. 1-4,doi: 10.1109/KICSS.2016.7951418. [3] Guodong Guo, S. Z. Li, and Kapluk Chan., \"Face recognition by support vector machines\", Proceedings Fourth IEEE International Conference on Automatic Face and Gesture Recognition, PP.196-201, doi: 10.1109/AFGR.2000.840634. [4] A. B. Perdana and A. Prahara., \"Face Recognition Using Light-Convolution Neural Networks Based On Modified Vgg16 Model”, International Conference of Computer ScienceandInformationTechnology.PP.1-4,doi:10.1109/ICoSNIKOM48755.2019.9111481. [5] A. Chowdhury and S. S. Tripathy., \"Human skin detection and face recognition using fuzzy logic and Eigen face”, International Conference on Green Computing Communication and Electrical Engineering, PP.1-4,doi:10.1109/ICGCCEE.2014.6922222. [6] D. M. M. da Costa, S. M. Peres, C. A. M. Lima, and P. Mustard., \"Face recognition using Support Vector Machine and multi scale directional image representation methods: A comparative study”, International Joint Conference on Neural Networks, PP.1-8,doi: 10.1109/IJCNN.2015.7280699. [7] Y. Chen, C. Liu, K. Chou, and S. Wang, \"Real-time and low-memory multi-face detection system design based on naive Bayes classifier using FPGA,\" International Automatic Control Conference (CACS), pp. 7-12, doi: 10.1109/CACS.2016.7973875. [8] H. Verma, S. Lotia and A. Singh., \"Convolution Neural Network Based Criminal Detection\", IEEERegion-10Conferences,PP.1124-1129,doi:10.1109/TENCON50793.2020.9293926. [9] M. M. Ghazi and H. K. Ekenel., \"A Comprehensive Analysis of Deep Learning Based Representation for Face Recognition\", IEEE Conference on Computer Vision and Pattern Recognition Workshops, PP.102-109, doi: 10.1109/CVPRW.2016.20. [10] I. K. Timotius, T. C. Linasari, I. Setyawan, and A. A. Febrianto., \"Face recognition using support vector machines and generalized discriminate analysis\", 6th International Conference onTelecommunicationSystems,ServicesandApplications,PP.8-10,doi:10.1109/TSSA.2011.6095397. [11] M. Nakada, H. Wang, and D. Terzopoulos., \"AcFR: Active Face Recognition Using Convolution Neural Networks,\" 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops, PP. 35-40, doi: 10.1109/CVPRW.2017.11. [12] Qurat-ul-ain, N. Nida, A. Irtaza and N.Ilyas., \"Forged Face Detection using ELA and Deep Learning Techniques\", International Bhurban Conference on Applied Sciences and Technologies, PP.271-275, doi: 10.1109/IBCAST51254.2021.9393234. [13] B. Heisele, P. Ho and T. Poggio, \"Face recognition with support vector machines: global versus component-based approach,\" Proceedings Eighth IEEE International Conference on Computer Vision, PP. 688-694 vol.2, doi: 10.1109/ICCV.2001.937693.

Copyright

Copyright © 2026 Dr. Tadi Chandrasekhar, Prof. Th. Basanta, Dr. Mutum Bidyarani Devi, Dr. J. N. Swaminathan. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.