Facial expression classification and morphing with machine learning
LE3 .A278 2020
2020
Silver, Danny
Acadia University
Master of Science
Masters
Computer Science
Automatic facial expression recognition and facial morphing are challenging problems in the field of machine learning and computer vision. In this thesis, we design and implement an automatic web-based system for facial expression recognition and facial morphing using machine learning with deep neural network and transfer learning. The website captures a photo desktop, laptop or phone, automatically detects a face, crops and pre-process the image, and sends it two machine learning models. One neural network model classifies the facial expression as being happiness, sadness, surprise, anger, fear, disgust or neutral. The second deep learning model transforms the pre-processed facial image into one of six expressive states, not included neutral, as requested by the user. Our system was tested on an independent set of images using a K-fold cross validation with a collection of images of 184 people. The proposed classification model achieved an F−measure of 94%. The morphing model achieved an accuracy of 89.64% on test data by comparing the original image and the generated image, using the Mean Absolute Error. The final website, face-morph.com, is capable of accurately classifying and morphing a facial image.
The author retains copyright in this thesis. Any substantial copying or any other actions that exceed fair dealing or other exceptions in the Copyright Act require the permission of the author.
https://scholar.acadiau.ca/islandora/object/theses:3544
