Evrişimsel Sinir Ağı Modellerine Dayalı Irk ve Cinsiyet Tanıma

Öz

Araştırmacılar ve geliştiriciler, şaşırtıcı ve hızlı bir ilerleme ile önyargı ve temsil sorunları da dahil olmak üzere birçok uygulamada derin öğrenmeyi ve bilgisayarlı görmeyi yaygın olarak kullanmaktadır. Geliştiriciler, insanları cinsiyet ve etnik kökene göre tanımlamak ve ayırt etmek için hem renk konsantrasyonunu hem de yüz ayrıntılarını kullanır. Bu makale, farklı ırklara ait yüz şekillerini tanımak için yeni bir evrişimli sinir ağı modeli kullanmaktadır. Model dört ırk ve cinsiyet (Afrikalı, Asyalı, Hintli ve Kafkasyalı) üzerinde eğitilerek test edilmiştir. Veri setlerinden toplanan veriler (pretty-face, SCUT-FBP5500_v2., AFD-dataset, cnsifd_faces_bmp, Indian_actors_faces, img_align_celeba, CASIA-Face-Africa). Elde edilen bulgular, Res50 modelinin testlerde daha iyi bir model doğruluk oranına sahip olduğunu göstermektedir. Testlerde Race Gender Convolution Neural Network (RGCNN) ve IncV3, Res50’den sonra en iyi doğruluk oranlarına sahiptir. VGG19 ise doğruluk oranlarında son sırada yer almaktadır. Cinsiyete göre hem Res50 hem de IncV3 daha iyi bir doğruluk oranına sahiptir. RGCNN üçüncü sırada, VGG19 ise son sırada yer almmaktadır. Parametre sayıları açısından değerlendirildiğinde RGCNN modeli en az parametre sayısına sahiptir. VGG19 Modeli ise ikinci sırada yer almaktadır. IncV3 modeli ise üçüncü sırada yer alırken son olarak Res50 modeli en çok parametre sayısına sahip olan modeldir.

Anahtar Kelimeler

• Convolutional Neural Network
• VGG19
• IncV3
• Resnet50
• Irk tanıma

Facial Race and Gender Recognition Based on Convolutional Neural Network Models

Öz

Researchers and developers have widely used deep learning and computer vision in many applications, including bias and representation issues, with amazing and rapid progress. In order to identify and differentiate people based on gender and ethnicity, developers employ both color concentration and facial details. This paper utilizes a new convolutional neural network model to recognize facial race. We trained and tested the model on four races and genders (African, Asian, Indian, and Caucasian). the dataset collected from the datasets (pretty-face, SCUT-FBP5500_v2., called AFD-dataset, cnsifd_faces_bmp, Indian_actors_faces, img_align_celeba, CASIA-Face-Africa). The experiment results show that Res50 models have proven to have a better model accuracy rate in races. Race Gender Convolution Neural Network (RGCNN) and IncV3 both achieved second place, while VGG19 ranked last. Both the Res50 and IncV3 results by gender show a better accuracy rate. RGCNN is in third place, while VGG19 is the last one. The RGCNN model is a lightweight has a smaller total number of parameters. The VGG19 Model, on the other hand, comes in second place. The IncV3 model, on the other hand, comes in third place, and finally, the Res50 model is the last one to have a total number of parameters.

Anahtar Kelimeler

• Facial Race Recognition
• Convolutional Neural Network
• VGG19
• IncV3
• Resnet50

Kaynakça

1. [1] Shaheed K, Szczuko P, Kumar M, Qureshi I , Abbas Q, and Ullah I J E A o A I. Deep learning techniques for biometric security: A systematic review of presentation attack detection systems: Engineering Applications of Artificial Intelligence. 2024; 129:107569.
2. [2] Ugale S, Patil W, Kapur V. IGRCVRM: Design of an Iterative Graph Based Recurrent Convolutional Model for Content Based Video Retrieval Using Multidomain Features, International Journal of Intelligent Systems and Applications in Engineering, 2024; 12(5s): 243-257.
3. [3] İNİK Ö and TURAN B. Classification of Different Age Groups of People by Using Deep Learnings: Dergipark. 2018; 7(3):9-16
4. [4] F. U. M. Ullah, M. S. Obaidat, A. Ullah, K. Muhammad, M. Hijji, and S. W. J. A. C. S. Baik, A comprehensive review on vision-based violence detection in surveillance videos, 2023; 55(10): 1-44.
5. [5] İNİK Ö and TURAN B. Classification of Animals with Different Deep Learning Models: Dergipark. 2018; 7(1):9-16.
6. [6] Tatar, A. B. "Biometric identification system using EEG signals.Neural Computing and Applications. 2023; 35(1): 1009-1023.
7. [7] Dargan S, Kumar M. A comprehensive survey on the biometric recognition systems based on physiological and behavioral modalities. Expert Systems with Applications.2020; 1(143):113114.
8. [8] Fitzgerald RJ, Price HL. Eyewitness identification across the life span: A meta-analysis of age differences. Psychoogicall Bulletin. 2015;141(6):1228-65.

9. [9] Li Z, Liu F, Yang W, Peng S, Zhou J. A survey of convolutional neural networks: analysis, applications, and prospects. IEEE transactions on neural networks and learning systems. 2021.
10. [10] Liu Y, Pu H, Sun DW. Efficient extraction of deep image features using convolutional neural network (CNN) for applications in detecting and analysing complex food matrices. Trends in Food Science & Technology. 2021 Jul 1;113:193-204.
11. [11] Young SG, Hugenberg K, Bernstein MJ, Sacco DF. Perception and motivation in face recognition: A critical review of theories of the cross-race effect. Personality and Social Psychology Review. 2012 May;16(2):116-42.
12. [12] Katti H, Arun SP. Can you tell where in India I am from? Comparing humans and computers on fine-grained race face classification. arXiv preprint arXiv:1703.07595. 2017 Mar 22.
13. [13] Puc A, Štruc V, Grm K. Analysis of race and gender bias in deep age estimation models. In2020 28th European Signal Processing Conference (EUSIPCO) 2021; Jan 18 pp. 830-834.
14. [14] Wang M, Deng W, Hu J, Tao X, Huang Y. Racial faces in the wild: Reducing racial bias by information maximization adaptation network. InProceedings of the ieee/cvf international conference on computer vision 2019 (pp. 692-702).
15. [15] Narang N, Bourlai T. Gender and ethnicity classification using deep learning in heterogeneous face recognition. In2016 International Conference on biometrics (ICB) 2016 Jun 13 (pp. 1-8).IEEE.
16. [16] Muhammad G, Hussain M, Alenezy F, Bebis G, Mirza AM, Aboalsamh H. Race recognition from face images using Weber local descriptor. In2012 19th International Conference on Systems, Signals and Image Processing (IWSSIP) 2012 Apr 11 (pp. 421-424). IEEE.
17. [17] Roomi SM, Virasundarii SL, Selvamegala S, Jeevanandham S, Hariharasudhan D. Race classification based on facial features. In2011 third national conference on computer vision, pattern recognition, image processing and graphics 2011 Dec 15 (pp. 54-57). IEEE.
18. [18]YEW T.S Pretty Face. https://www.kaggle.com/datasets/yewtsing/pretty-face.
19. [19] HCIILAB. SCUT-FBP5500 Database Release. https://github.com/HCIILAB/SCUT
20. [20] ZHANG Y. Asian Face Image Dataset (AFD) [Dataset]. GitHub.https://github.com/X-zhangyang/Asian-Face-Image-Dataset-AFD-dataset
21. [21] HARISH. IISCIFD [Dataset]. GitHub. https://github.com/harish2006/IISCIFD
22. [22]NAGASAI524.IndianActorFacesforFaceRecognition[Dataset]. https://www.kaggle.com/datasets/nagasai524/indian-actor-faces-for-face-recognition
23. [23]MMLAB.CelebA:Large-ScaleCelebFacesAttributesDataset [Dataset].http://mmlab.ie.cuhk.edu.hk/projects/CelebA.html
24. [24]PAPERSWITHCODE.CASIAFaceAfrica[Dataset].https://paperswithcode.com/dataset/casia-face-africa
25. [25] Agbo-Ajala O, Viriri S. Deep learning approach for facial age classification: a survey of the state-of-the-art. Artificial Intelligence Review. 2021 Jan;54:179-213.
26. [26] Ruan X, Liu Y, Yuan C, Li B, Hu W, Li Y, Maybank S. Edp: An efficient decomposition and pruning scheme for convolutional neural network compression. IEEE Transactions on Neural Networks and Learning Systems. 2020 Nov 2;32(10):4499-513
27. [27] Sam SM, Kamardin K, Sjarif NN, Mohamed N. Offline signature verification using deep learning convolutional neural network (CNN) architectures GoogLeNet inception-v1 and inception-v3. Procedia Computer Science. 2019 Jan 1;161:475-83.
28. [28] Alom MZ, Hasan M, Yakopcic C, Taha TM, Asari VK. Improved inception-residual convolutional neural network for object recognition. Neural Computing and Applications. 2020 Jan;32:279-93.
29. [29] Jena B, Nayak GK, Saxena S. Convolutional neural network and its pretrained models for image classification and object detection: A survey. Concurrency and Computation: Practice and Experience. 2022 Mar 10;34(6):e6767.
30. [30] Zhang L, Bian Y, Jiang P, Zhang F. A Transfer Residual Neural Network Based on ResNet-50 for Detection of Steel Surface Defects. Applied Sciences. 2023 Apr 23;13(9):5260.
31. [31] Ikechukwu AV, Murali S, Deepu R, Shivamurthy RC. ResNet-50 vs VGG-19 vs training from scratch: A comparative analysis of the segmentation and classification of Pneumonia from chest X-ray images. Global Transitions Proceedings. 2021 Nov 1;2(2):375-81.
32. [32] Hindarto D. Revolutionizing Automotive Parts Classification Using InceptionV3 Transfer Learning. International Journal Software Engineering and Computer Science (IJSECS). 2023 Dec 10;3(3):324-33.
33. [33] Anand V, Gupta S, Koundal D, Mahajan S, Pandit AK, Zaguia A. Deep Learning Based Automated Diagnosis of Skin Diseases Using Dermoscopy. Computers, Materials & Continua. 2022 May 1;71(2).
34. [34] Xiang W, Rao H, Zhou L. A meta‐analysis of contrast‐enhanced spectral mammography versus MRI in the diagnosis of breast cancer. Thoracic cancer. 2020 Jun;11(6):1423-32.