Abstract:
Automated Emotion Extraction has always been an interesting research area. Now with the
advent of selfies (A form of frontal facial image based photography) and a mainstream
interest in organizing countless amount of pictures people take with the devices they own, the
practical applications of such heavily research based topics are increasing. In the light of
this, this project aims to build an efficient emotion extraction module which would target to
classify the photos taken based on the six basic expressions stated by Ekman and Friesen
namely happiness, sadness, anger, surprise, disgust, fear along with neutral and categorize
them under their respective groups. To demonstrate this module an Intelligent Digital Photo
Organizer will be built incorporating the above module which will organize the photos
imported to the application under above mentioned categories. The emotion extraction
module will be made in such a way that it could be seamlessly incorporated for other similar
purposes as well. Through an in-depth literature study, the system proposes two novel ways
of solving the above problem one with a traditional approach and the other with a deep
learning approach, latter achieving an average accuracy of 76.88 %.
