Official code repository for "Gender recognition in masked facial images using EfficinetNet and transfer learning approach"
The rapid advancement of Artificial Intelligence (AI) technology has enabled diverse systems to leverage individual characteristics to enhance their functionality. Among these attributes, gender information is pivotal in human-machine interaction across various domains, such as vending machines and targeted advertising campaigns. While numerous methodologies have been developed for gender detection from facial images, they often encounter limitations, particularly in scenarios involving masked individuals during unprecedented events like the COVID-19 pandemic. This study tries to present an optimal Convolutional Neural Network architecture for gender recognition based on EfficientNet, which is recognized as the most effective backbone network for gender recognition during some experiments, and train it with a masked-worn faces database to yield an efficient network for detecting the gender of masked-worn people. However, creating such datasets proves to be both time-consuming and resource-intensive. Using the Poisson Image Editing technique, this article introduces an innovative and cost-efficient approach to generating masked facial images from pre-existing face databases to surmount this obstacle. Remarkably, the accuracy achieved by the proposed methodologies on three renowned databases—LFW, CelebA, and Adience— for original images is an impressive 98.5%, 98.27%, and 95.80%, respectively. This outstanding performance demonstrates a significant advancement over prior endeavors, underscoring the efficiency and robustness of our approach.
- Clone codes.
git clone https://github.com/FaezehMosayyebi/Gender_Recognition.git
- Install requirements.
python3 -m venv .venv
.\.venv\Scripts\activate
pip install -r requirements.txt
- Run interaction.ipynb
When you publish your research using these codes, please cite [1] as
@article{
title = {Gender recognition in masked facial images using EfficientNet and transfer learning approach,
author = {Mosayyebi, Faezeh, Seyedarabi, Hadi, and Afrouzian, Reza},
journal = {International Journal of Information Technology},
volume = {16},
pages = {2693–2703},
year = {2024},
doi = {10.1007/s41870-023-01565-4},
publisher = {Springer},}
[1] Mosayyebi, F., Seyedarabi, H. & Afrouzian, R. Gender recognition in masked facial images using EfficientNet and transfer learning approach. Int. j. inf. tecnol. 16, 2693–2703 (2024). https://doi.org/10.1007/s41870-023-01565-4
Contact: faezeh.mosayyebi@gmail.com