A Deep Learning Approach To Face Shape Classification For Hairstyle, Beauty And Eyewear Recommendations

Accurate face shape classification plays an important role in the growing field of personalized recommendations for hairstyle, beauty, and eyewear designs. Face shape classification is a computer vision task which aims to identify the shape of a human face from an image. This paper introduces state-of-the-art deep learning approaches for accurate face shape classification using advanced convolutional neural networks (CNNs) such as GoogLeNet, MobileNetV3, ResNet50, and Sequential CNN. The system analyses images of faces with varying shapes, such as - oval, round, square, heart, and oblong and provides personalized recommendations. The results of the study shows that GoogLeNet achieves the highest accuracy at ${9 1 . 0 3 \%}$, followed by MobileNetV3 at ${8 9 . 7 4 \%}$, ResNet50 at ${8 8 . 5 4 \%}$, and the Sequential CNN at ${8 0 . 0 2 \%}$. The superior performance of GoogLeNet indicates its efficacy in capturing the intricate features necessary for precise face shape classification. Additionally, a Flask-based web application has been developed, which enables real-time face shape classification and provides personalized recommendations.