ADVANCED FACE DETECTION USING RESNET AND FPN ARCHITECTURES WITH FOCAL LOSS FOR ENHANCED ACCURACY

Authors

  • Agzamova Mohinabonu PhD student, Tashkent University of Information Technologies named after Muhammad al-Khwarizmi, Tashkent, Uzbekistan

Keywords:

Face detection, neural networks, Focal Loss

Abstract

This paper explores an innovative approach to face detection using neural networks based on the ResNet architecture and Feature Pyramid Network (FPN). The focus is on applying the Focal Loss function, which significantly improves classification accuracy for difficult examples while addressing class imbalance issues in face detection tasks. The paper discusses the stages of data preparation, including augmentation and preprocessing, and highlights the development and training of the model based on deep convolutional neural networks (CNNs). The paper demonstrates the high accuracy and recall of the algorithm on test datasets.

References

T.-Y. Lin et al., "Feature Pyramid Networks for Object Detection," CVPR, 2017.

P. Viola and M. Jones, "Robust Real-Time Face Detection," 2001.

D. Irgasheva et al., "Face Detection in Payment Systems," 2023.

P. J. Phillips et al., "The FERET Evaluation," 2018.

Downloads

Published

2024-10-30

How to Cite

Agzamova Mohinabonu. (2024). ADVANCED FACE DETECTION USING RESNET AND FPN ARCHITECTURES WITH FOCAL LOSS FOR ENHANCED ACCURACY. Next Scientists Conferences, 1(01), 48–51. Retrieved from https://nextscientists.com/index.php/science-conf/article/view/345

Issue

2024: EXAMINING THE CROSSROADS OF HISTORY, EDUCATION, AND SOCIETY: THEORY, PRACTICE, AND POLICY

Section

Articles