Facial Landmark Localization is the detection and localization of certain key points in a person’s face. This plays an important role in subsequent facial processing operations that range from biometric recognition to understanding a person’s emotional/mental state. This is a challenging problem in computer vision due to there being a multitude of factors constituting the placement of these key points on a human face such as expression, facial profile, pose, illumination, etc.
The method we used to accomplish the task of facial landmark detection was to use a Convolutional Neural Network (CNN). CNN is an artificial intelligence neural network that mainly comes into play with images. The dataset we used for this project was taken from the Kaggle platform. The dataset contained 5770 facial images and corresponding key points marked on them which represent the facial attributes of the eyes, mouth, and nose.
Each predicted keypoint was specified by an (x,y) real-valued pair in the space of pixel indices.