Both models are trained with over 3000 images of faces from the imdb-wiki dataset. The gender model is a simple binary
classification model and the age model is a regression model. They both fine-tune a DenseNet convolutional neural network.
The gender prediction model has an accuracy of about 92%, while the age prediction model has a mean absolute error of about 8.
Both models work better when you are closer to the camera and showing your full face and hair.