Summary

This code performs gender recognition based on voice using deep learning. The voice dataset is loaded and preprocessed by normalizing the data and splitting it into training and test sets. A sequential model with three dense layers and two dropout layers is built using TensorFlow. The model is then compiled with binary crossentropy loss function and Adam optimizer. After training the model with the training data, it is evaluated using the test data, and the accuracy and loss are printed. The model is also saved in an .h5 file. Furthermore, the saved model can be loaded, recompiled with a new optimizer and learning rate, trained with new data, and evaluated for performance. Finally, a plot of the training and validation accuracy is displayed. With an accuracy of 98.26% on the test data, this model can effectively predict gender based on voice.