Skip to main content
SpringerLink
Log in

Uncovering Bias in the Face Processing Pipeline: An Analysis of Popular and State-of-the-Art Algorithms Across Demographic Groups

  • Conference paper
  • First Online:
AIxIA 2023 – Advances in Artificial Intelligence (AIxIA 2023)

Part of the book series: Lecture Notes in Computer Science ((LNAI,volume 14318))

  • 446 Accesses

Abstract

Numerous algorithms process face images to perform tasks such as person identification and estimation of attributes such as the race and gender. While previous work has focused on biases in face recognition systems, relatively limited work has considered the full face processing pipeline to determine if other components also exhibit any biases related to a person’s demographic attributes. An evaluation of popular and state-of-the-art methods in the face processing pipeline reveals that, although the overall performance may appear satisfactory, numerous differences are uncovered when digging deeper to consider the performance not just within a single demographic group, but also across different types of groups. Several avenues of future work are also provided.

This is a preview of subscription content, log in via an institution to check access.

Access this chapter

Log in via an institution
Chapter
USD 29.95
Price excludes VAT (USA)
  • Available as PDF
  • Read on any device
  • Instant download
  • Own it forever
eBook
USD 59.99
Price excludes VAT (USA)
  • Available as EPUB and PDF
  • Read on any device
  • Instant download
  • Own it forever
Softcover Book
USD 79.99
Price excludes VAT (USA)
  • Compact, lightweight edition
  • Dispatched in 3 to 5 business days
  • Free shipping worldwide - see info

Tax calculation will be finalised at checkout

Purchases are for personal use only

Institutional subscriptions

Notes

  1. 1.

    Available at: https://github.com/serengil/deepface.

  2. 2.

    A facial recognition library that recognises and manipulates faces. Available at https://github.com/ageitgey/face_recognition.

  3. 3.

    Also known as a query set.

  4. 4.

    Available at: https://github.com/rcmalli/keras-vggface.

References

  1. How many major races are there in the world? (2011). https://blog.world-mysteries.com/science/how-many-major-races-are-there-in-the-world/

  2. Face recognition (2018). https://github.com/ageitgey/face_recognition/

  3. Badave, H., Kuber, M.: Face recognition based activity detection for security application. In: 2021 International Conference on Artificial Intelligence and Smart Systems (ICAIS), pp. 487–491 (2021). https://doi.org/10.1109/ICAIS50930.2021.9395829

  4. Blakemore, E.: Race and ethnicity: How are they different? https://www.nationalgeographic.com/culture/article/race-ethnicity (2019)

  5. Bradski, G.: The OpenCV library. Dr. Dobb’s J. Softw. Tools 25(11), 120–123 (2000)

    Google Scholar 

  6. Cao, Q., Shen, L., Xie, W., Parkhi, O.M., Zisserman, A.: VGGFace2: a dataset for recognising faces across pose and age. In: 2018 13th IEEE International Conference on Automatic Face & Gesture Recognition (FG 2018), pp. 67–74. IEEE Computer Society, Los Alamitos, CA, USA (2018). https://doi.org/10.1109/FG.2018.00020

  7. Cascone, L., Pero, C., Proença, H.: Visual and textual explainability for a biometric verification system based on piecewise facial attribute analysis. Image Vis. Comput. 132, 104645 (2023)

    Article  Google Scholar 

  8. Cavazos, J.G., Phillips, P.J., Castillo, C.D., O’Toole, A.J.: Accuracy comparison across face recognition algorithms: where are we on measuring race bias? IEEE Trans. Biometrics Behav. Identity Sci. 3(1), 101–111 (2021). https://doi.org/10.1109/TBIOM.2020.3027269

    Article  Google Scholar 

  9. Das, A., Dantcheva, A., Bremond, F.: Mitigating bias in gender, age and ethnicity classification: a multi-task convolution neural network approach. In: Proceedings of the European Conference on Computer Vision (ECCV) Workshops (2018)

    Google Scholar 

  10. Deng, J., Guo, J., Xue, N., Zafeiriou, S.: ArcFace: additive angular margin loss for deep face recognition. In: 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 4685–4694 (2019). https://doi.org/10.1109/CVPR.2019.00482

  11. Drozdowski, P., Rathgeb, C., Dantcheva, A., Damer, N., Busch, C.: Demographic bias in biometrics: a survey on an emerging challenge. IEEE Trans. Technol. Soc. 1(2), 89–103 (2020). https://doi.org/10.1109/TTS.2020.2992344

    Article  Google Scholar 

  12. Du, H., Shi, H., Zeng, D., Zhang, X.P., Mei, T.: The elements of end-to-end deep face recognition: a survey of recent advances. ACM Comput. Surv. (CSUR) 54(10s), 1–42 (2022)

    Article  Google Scholar 

  13. Ferrari, C., Lisanti, G., Berretti, S., Del Bimbo, A.: Investigating nuisances in DCNN-based face recognition. IEEE Trans. Image Process. 27(11), 5638–5651 (2018). https://doi.org/10.1109/TIP.2018.2861359

    Article  MathSciNet  MATH  Google Scholar 

  14. Furl, N., Phillips, P., O’Toole, A.J.: Face recognition algorithms and the other-race effect: computational mechanisms for a developmental contact hypothesis. Cogn. Sci. 26(6), 797–815 (2002)

    Article  Google Scholar 

  15. Galea, C., Farrugia, R.A.: Forensic face photo-sketch recognition using a deep learning-based architecture. IEEE Signal Process. Lett. 24(11), 1586–1590 (2017). https://doi.org/10.1109/LSP.2017.2749266

    Article  Google Scholar 

  16. Galea, C., Farrugia, R.A.: Matching software-generated sketches to face photographs with a very deep CNN, morphed faces, and transfer learning. IEEE Trans. Inf. Forensics Secur. 13(6), 1421–1431 (2018). https://doi.org/10.1109/TIFS.2017.2788002

    Article  Google Scholar 

  17. Galea, N., Seychell, D.: Facial expression recognition in the wild: dataset configurations. In: 2022 IEEE 5th International Conference on Multimedia Information Processing and Retrieval (MIPR), pp. 216–219 (2022). https://doi.org/10.1109/MIPR54900.2022.00045

  18. Georgopoulos, M.: Bias in deep learning and applications to face analysis. Ph.D. thesis, Imperial College London (2022)

    Google Scholar 

  19. Gong, S., Liu, X., Jain, A.K.: Jointly de-biasing face recognition and demographic attribute estimation. In: Vedaldi, A., Bischof, H., Brox, T., Frahm, J.-M. (eds.) ECCV 2020. LNCS, vol. 12374, pp. 330–347. Springer, Cham (2020). https://doi.org/10.1007/978-3-030-58526-6_20

    Chapter  Google Scholar 

  20. He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 770–778 (2016)

    Google Scholar 

  21. Hu, J., Shen, L., Sun, G.: Squeeze-and-excitation networks. In: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 7132–7141 (2018). https://doi.org/10.1109/CVPR.2018.00745

  22. Karkkainen, K., Joo, J.: FairFace: face attribute dataset for balanced race, gender, and age for bias measurement and mitigation. In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pp. 1548–1558 (2021)

    Google Scholar 

  23. King, D.E.: Dlib-ml: a machine learning toolkit. J. Mach. Learn. Res. 10, 1755–1758 (2009)

    Google Scholar 

  24. Klare, B.F., et al.: Pushing the frontiers of unconstrained face detection and recognition: iarpa janus benchmark a. In: 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 1931–1939 (2015). https://doi.org/10.1109/CVPR.2015.7298803

  25. Malli, R.C.: keras-vggface (2019). https://github.com/rcmalli/keras-vggface

  26. Mather, M., Jacobsen, L.A., Scommegna, P.: Population: an introduction to demography. Popul. Bull. 75(1) (2021)

    Google Scholar 

  27. Maze, B., et al.: Iarpa janus benchmark - c: face dataset and protocol. In: 2018 International Conference on Biometrics (ICB), pp. 158–165 (2018). https://doi.org/10.1109/ICB2018.2018.00033

  28. Neto, P.C., et al.: Explainable biometrics in the age of deep learning (2022)

    Google Scholar 

  29. Parkhi, O.M., Vedaldi, A., Zisserman, A.: Deep face recognition. In: British Machine Vision Conference (2015)

    Google Scholar 

  30. Pedregosa, F., et al.: Scikit-learn: machine learning in Python. J. Mach. Learn. Res. 12, 2825–2830 (2011)

    MathSciNet  MATH  Google Scholar 

  31. Phillips, P., Wechsler, H., Huang, J., Rauss, P.J.: The FERET database and evaluation procedure for face-recognition algorithms. Image Vis. Comput. 16(5), 295–306 (1998). https://doi.org/10.1016/S0262-8856(97)00070-X

    Article  Google Scholar 

  32. Popa, M., Rothkrantz, L., Yang, Z., Wiggers, P., Braspenning, R., Shan, C.: Analysis of shopping behavior based on surveillance system. In: 2010 IEEE International Conference on Systems, Man and Cybernetics, pp. 2512–2519. IEEE (2010)

    Google Scholar 

  33. Ribeiro, M.T., Singh, S., Guestrin, C.: “Why should i trust you?”: explaining the predictions of any classifier. In: Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD 2016, pp. 1135–1144. Association for Computing Machinery, New York, NY, USA (2016). https://doi.org/10.1145/2939672.2939778

  34. Robinson, J.P., Livitz, G., Henon, Y., Qin, C., Fu, Y., Timoner, S.: Face recognition: too bias, or not too bias? In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (2020)

    Google Scholar 

  35. Rodríguez Spinelli, F.: Facing the bias: when face processing algorithms bump into diversity. SSRN 3823012 (2021)

    Google Scholar 

  36. Rothe, R., Timofte, R., Van Gool, L.: DEX: deep expectation of apparent age from a single image. In: 2015 IEEE International Conference on Computer Vision Workshop (ICCVW), pp. 252–257 (2015). https://doi.org/10.1109/ICCVW.2015.41

  37. Selvaraju, R.R., Cogswell, M., Das, A., Vedantam, R., Parikh, D., Batra, D.: Grad-CAM: visual explanations from deep networks via gradient-based localization. In: 2017 IEEE International Conference on Computer Vision (ICCV), pp. 618–626 (2017). https://doi.org/10.1109/ICCV.2017.74

  38. Serengil, S.I.: Apparent age and gender prediction in keras (2019). https://sefiks.com/2019/02/13/apparent-age-and-gender-prediction-in-keras/

  39. Serengil, S.I., Ozpinar, A.: Lightface: a hybrid deep face recognition framework. In: 2020 Innovations in Intelligent Systems and Applications Conference (ASYU), pp. 23–27. IEEE (2020). https://doi.org/10.1109/ASYU50717.2020.9259802

  40. Serengil, S.I., Ozpinar, A.: Hyperextended lightface: a facial attribute analysis framework. In: 2021 International Conference on Engineering and Emerging Technologies (ICEET), pp. 1–4. IEEE (2021). https://doi.org/10.1109/ICEET53442.2021.9659697

  41. Simonyan, K., Zisserman, A.: Very deep convolutional networks for large-scale image recognition. In: Bengio, Y., LeCun, Y. (eds.) 3rd International Conference on Learning Representations, ICLR 2015, San Diego, May 7–9 2015, Conference Track Proceedings (2015). https://arxiv.org/abs/1409.1556

  42. Terhörst, P., et al.: Reliable age and gender estimation from face images: stating the confidence of model predictions. In: 10th IEEE International Conference on Biometrics Theory, Applications and Systems, BTAS 2019, Tampa, FL, USA, 23–26 September 2019, pp. 1–8. IEEE (2019). https://doi.org/10.1109/BTAS46853.2019.9185975

  43. Terhörst, P., et al.: A comprehensive study on face recognition biases beyond demographics. IEEE Trans. Technol. Soc. 3(1), 16–30 (2022). https://doi.org/10.1109/TTS.2021.3111823

    Article  Google Scholar 

  44. Thom, N., Hand, E.M.: Facial attribute recognition: a survey. Comput. Vis.: Ref. Guide, 1–13 (2020)

    Google Scholar 

  45. Whitelam, C., et al.: Iarpa janus benchmark-b face dataset. In: 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), pp. 592–600 (2017). https://doi.org/10.1109/CVPRW.2017.87

  46. Won, D.: Gender and race classification with face images (2018). https://github.com/wondonghyeon/face-classification

  47. Zhang, K., Zhang, Z., Li, Z., Qiao, Y.: Joint face detection and alignment using multitask cascaded convolutional networks. IEEE Signal Process. Lett. 23(10), 1499–1503 (2016). https://doi.org/10.1109/LSP.2016.2603342

    Article  Google Scholar 

Download references

Acknowledgements

This research was supported by the project REtics (Review Analytics), which is financed by the Malta Council for Science & Technology, for and on behalf of the Foundation for Science and Technology, through the FUSION: R &I Technology Development Programme LITE.

Author information

Authors and Affiliations

  1. Beyond Museums R &D, Birkirkara, Malta

    Christian Galea, Chantelle Saliba, Matthew Sacco & Mark Bugeja

  2. SeyTravel Ltd, Msida, Malta

    Christian Galea, Chantelle Saliba, Matthew Sacco & Mark Bugeja

  3. Institute for Tourism and Culture, University of Malta, Msida, Malta

    Noel Buttigieg

  4. Department of Artificial Intelligence, University of Malta, Msida, Malta

    Dylan Seychell

Authors
  1. Christian Galea
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Chantelle Saliba
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Matthew Sacco
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Mark Bugeja
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Noel Buttigieg
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Dylan Seychell
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Christian Galea .

Editor information

Editors and Affiliations

  1. University of Rome Tor Vergata, Rome, Italy

    Roberto Basili

  2. Sapienza University of Rome, Rome, Italy

    Domenico Lembo

  3. Roma Tre University, Rome, Italy

    Carla Limongelli

  4. National Research Council, Rome, Italy

    Andrea Orlandini

A Appendix: Supplementary Information

A Appendix: Supplementary Information

This appendix contains additional images and tables that supplement the discussion in the main manuscript, as follows:

  • Table 5 and Table 6 show the performance of the gender and race estimation methods, respectively, across all subjects.

  • Table 7 contains the confidence scores for the race estimation methods, for each race considered.

  • Table 8 and Table 9 contain the performance of the FRSs when evaluated on male and female subjects, respectively.

  • Fig. 5 and Fig. 6 show the accuracy scores of the methods considered over different racial groups for male and female subjects, respectively

  • Fig. 7 depicts examples of an incorrect match by OpenCV + VGG-Face (VGG16) [DF].

  • Fig. 8 shows the confusion matrices for the gender and race estimation methods.

Table 5. Performance of the gender estimation methods, across all subjects. ‘C’ and ‘I’ refer to the mean confidence values output by the method when the prediction is correct and when the prediction is incorrect, respectively. ‘A’ represents the confidence score of the correct label when the prediction is incorrect. All values in %.
Full size table
Table 6. Performance of the race estimation methods, across all subjects. ‘C’ and ‘I’ refer to the mean confidence values output by the method when the prediction is correct and when the prediction is incorrect, respectively. ‘A’ represents the confidence score of the correct label when the prediction is incorrect. All values in %.
Full size table
Table 7. Confidence scores for race estimation methods, for each race. ‘C’ and ‘I’ denote confidence scores when the predicted label is correct and when the predicted label is incorrect, respectively. ‘A’ represents the confidence score of the correct label when the prediction is incorrect. All values in %.
Full size table
Table 8. FRS performance for ‘Male’ subjects in each demographic group. [DF] denotes that the method as implemented in the DeepFace library. The rest of the methods use the implementations in [25]. ‘Acc.’ denotes ‘Accuracy’.
Full size table
Table 9. FRS performance for ’Female’ subjects in each demographic group. [DF] denotes that the method as implemented in the DeepFace library. The rest of the methods use the implementations in [25]. ‘Acc.’ denotes ‘Accuracy’.
Full size table
Fig. 5.
figure 5

Accuracy scores of the methods considered over different racial groups for male subjects

Full size image
Fig. 6.
figure 6

Accuracy scores of the methods considered over different racial groups for female subjects

Full size image
Fig. 7.
figure 7

Example of an incorrect match by OpenCV + VGG-Face (VGG16) [DF]. Three of the top five subjects wear glasses, like the query subject.

Full size image
Fig. 8.
figure 8

Normalised confusion matrices for the gender and race estimation methods. Labels the same as those used by the methods considered.

Full size image

Rights and permissions

Reprints and permissions

Copyright information

© 2023 The Author(s), under exclusive license to Springer Nature Switzerland AG

About this paper

Check for updates. Verify currency and authenticity via CrossMark

Cite this paper

Galea, C., Saliba, C., Sacco, M., Bugeja, M., Buttigieg, N., Seychell, D. (2023). Uncovering Bias in the Face Processing Pipeline: An Analysis of Popular and State-of-the-Art Algorithms Across Demographic Groups. In: Basili, R., Lembo, D., Limongelli, C., Orlandini, A. (eds) AIxIA 2023 – Advances in Artificial Intelligence. AIxIA 2023. Lecture Notes in Computer Science(), vol 14318. Springer, Cham. https://doi.org/10.1007/978-3-031-47546-7_17

Download citation

  • DOI: https://doi.org/10.1007/978-3-031-47546-7_17

  • Published:

  • Publisher Name: Springer, Cham

  • Print ISBN: 978-3-031-47545-0

  • Online ISBN: 978-3-031-47546-7

  • eBook Packages: Computer ScienceComputer Science (R0)

Publish with us

Policies and ethics

Search

Navigation