Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

Exploring the Feasibility of Deep Learning Techniques for Accurate Gender Classification from Eye Images

Created by
  • Haebom

Author

Basna Mohammed Salih Hasan, Ramadhan J. Mstafa

Outline

This paper presents a gender classification model using color images of the periocular region, unaffected by factors such as makeup or disguise. The proposed CNN model was evaluated on two eye datasets: CVBL and (Female and Male). It achieved high accuracies of 99% on the CVBL dataset and 96% on the (Female and Male) dataset. This was achieved using a small number of learnable parameters (7,235,089). The model's performance was evaluated using various metrics and compared to existing state-of-the-art techniques, demonstrating its effectiveness and suggesting practical applications in areas such as security and surveillance.

Takeaways, Limitations

•
Takeaways:
◦
Demonstrating the effectiveness of gender classification using the area around the eyes.
◦
We present an efficient CNN model that achieves high accuracy (96-99%).
◦
Increase computational load and resource efficiency by implementing high performance with fewer parameters.
◦
Presenting practical applications in various fields, including security and surveillance.
•
Limitations:
◦
Lack of specific details about the size and diversity of the datasets used.
◦
Lack of generalization performance assessment across different races or age groups.
◦
Further validation of performance and robustness in real-world application environments is required.
◦
(Female and Male) Lack of specific information in the dataset.
👍
Made with Slashpage