Emotion Detection Using Facial Landmarks

B. Dwarakanath, S.K. Surya, A.R. Nandu Krishna and P. Lava Kumar

Traditionally, when a computer system needs perform a job, a set of instructions is given to perform the job, known as a computer program. This program is written extensively by a human, and heavily relied on human logic to do so. But, the problem is that, there are tasks that are too complex for humans to tell the computer every single detail of how to do a task. Thus, a method is coined for the computers to learn by themselves on how to do a task, Known as ā€œArtificial Intelligenceā€. There have been many attempts to teach a computer to try and extract the emotion from an image of the face. Conventional wisdom will suggest that we simply feed thousands of images of various emotional states to conventional neural networks, and are also the limited too. Hence, a machine learning system is used to solve this problem with test and train data for system to learn about the emotion. The maximum test set accuracy detecting emotion achieved previously on the fer2013 dataset is 60%. This paper aims at improving the emotion detection closer to 100% and displays the emotions of the person reflected on his face.

Volume 11 | 06-Special Issue

Pages: 148-155