Methodologies & Tools

Programming Language | Python
Methodologies | Lasso, PCA, Forward-Backward Algorithm, Deep Learning, VGG

Introduction
The problem and the data come from the Kaggle online competition. The data consists of a list of 7049 2-D and 8-bit gray level training images with their corresponding (x; y) coordinates of the 15 facial keypoints for eyes, eyebrows, nose and mouth. Each image is represented by a 96 x 96- pixel matrix with integers representing pixel value from 0 through 255, there are in total 9216 pixels. After cleaning and train/test splitting, 1712 images are used as the train data, and 428 images are use as the test data for all the methods.