Investigating Bias in Gender Classification Models

A Convolutional Neural Networks (CNN) is a type of deep learning algorithm which is often used for image classification tasks. One such image classification task is the classification of a person’s gender based upon an image. These gender classification models often pose a bias towards transgender and dark skinned individuals. Such biases are compounded by the fact that large neural networks function as black boxes and lack interpretability; it is difficult to explain why a model classifies an image in a particular way. We investigate Gradient-weighted Class Activation Mapping (Grad-CAM) as a visualization tool which can explain how gender classification models make their decisions.