Motivation
Detecting facial expression promotes communication between individuals. For humans, recognizing facial expressions and emotions is a basic skill that is learned at an early age. We can look at a person’s face and can quickly recognize the common emotions of anger, happiness, surprise, disgust, sadness, and fear. However, it is challenging to transfer this skill to a machine is a complex task. Despite of its complexity, automatic real-time detection on humans’ facial expressions can potentially be an useful computer vision technology in real life. Some possible practical applications are given below.
- The US Department of Transportation claims that driving-related errors cause around 95 percent fatal road accidents. Facial emotion detection can help to find changes in drivers’ facial expressions which precede drowsiness or anger that are harmful to safe driving, so that personalized alerts can be sent to the drivers in time.
- In business, customers’ facial expressions often reflect their feelings on the current ongoing service. Detection on their facial expressions can help companies in both collecting customers’ feedbacks effectively and altering the service in time, which then can improve customers’ experience and protect companies’ benefits.
- In daily life, at each time when people open their computers or cellphones, their moods are possibly reflected on their faces. Detecting their facial expressions and sending customized greetings or suggestions can make them feel even better or help them improve their emotions.
As these examples have revealed, real-time detection on facial expressions can be helpful in many real-life applications. Accuracy of detection becomes important in practical practices. Currently, deep learning neural networks appear to be a good tool in increasing accuracy of classification tasks. This meets our goal as we want to accurately classify a given image into a specific facial expression. Therefore, we are interested in finding a deep learning model that can give accurate results as well as can be easily applied to a real-time detection system.