Age Estimation from Facial Images using Machine Learning
Nguyen, Dang Chau (2023)
Nguyen, Dang Chau
2023
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2023060923264
https://urn.fi/URN:NBN:fi:amk-2023060923264
Tiivistelmä
In today's era of online services, there is a growing need to verify user age to comply with legal requirements. However, traditional methods requiring collecting personal identification data may compromise user privacy and security. In response to the challenge, this thesis presents an approach based on machine learning for age estimation, designed to predict user age from facial images without collecting all sensitive user data. As a result, this proposed approach reduces the resources required to process and store sensitive user information, reducing the potential impact of data breaches.
The proposed approach uses Machine Learning models, such as ResNet and MobileNet, to process facial images and predict user age. The approach is evaluated using various performance metrics. Despite initial promising results, the proposed approach did not reach the desired accuracy due to overfitting issues, signaling the need for further refinement and research. Nonetheless, the approach promises reduced resources needed to process private information, potentially lowering significant costs, and offering efficiency benefits for online services.
In conclusion, while the proposed machine learning based age verification approach showed potential, it did not fully achieve the expected outcomes in this study, underlining the need for continued investigation. Future research should also consider potential applications in other forms of biometric identification while prioritizing user privacy and security.
The proposed approach uses Machine Learning models, such as ResNet and MobileNet, to process facial images and predict user age. The approach is evaluated using various performance metrics. Despite initial promising results, the proposed approach did not reach the desired accuracy due to overfitting issues, signaling the need for further refinement and research. Nonetheless, the approach promises reduced resources needed to process private information, potentially lowering significant costs, and offering efficiency benefits for online services.
In conclusion, while the proposed machine learning based age verification approach showed potential, it did not fully achieve the expected outcomes in this study, underlining the need for continued investigation. Future research should also consider potential applications in other forms of biometric identification while prioritizing user privacy and security.