Real-Time Gender Classification for Mizo Community using Convolutional Neural Networks

Real-time facial classification for gender identification is crucial in various applications, such as restroom access control, where authentication is based on the confidence score from each facial image. This study focuses on gender classification within the Mizo community, using CNN to address the unique facial features characteristic of this group. The main goal is to develop an accurate model for the Mizo community. Numerous male and female facial images were collected, and augmentation techniques were applied. Preprocessing involved converting images to grayscale and using Principal Component Analysis (PCA) for dimensionality reduction. The CNN model, built on MobileNet, was trained with K-fold cross-validation, consistently achieving high performance. The model is deployed for real-time classification, accurately identifying gender with confidence scores. It demonstrates $98 \%$ to $99 \%$ accuracy, proving its robustness and effectiveness in real-world applications.