• home
• current students
• graduate
• events

Presenter: Mohammed Adnan Adil
Supervisor:

Date: Thu, September 2, 2021
Time: 10:00:00 - 11:00:00
Place: via Zoom - please see link below

ABSTRACT

Zoom meeting link: 

https://uvic.zoom.us/j/8537457353?pwd=OXBJWXR0VGFWNkw1bHBXaUZjOW5jQT09

Meeting ID: 853 745 7353

Password: 729166

One tap mobile

+17789072071,,8537457353# Canada

+16475580588,,8537457353# Canada

Dial by your location

        +1 778 907 2071 Canada

        +1 647 558 0588 Canada

Meeting ID: 853 745 7353

Find your local number: https://uvic.zoom.us/u/kexEilF5Df

Note: Please log in to Zoom via SSO and your UVic Netlink ID

ABSTRACT:

Human emotions are the mental state of feelings and are spontaneous. There is no clear connection between emotions and facial expressions and there is significant variability making facial recognition a challenging research area. Features like Histogram of Oriented Gradient (HOG) and Scale Invariant Feature Transform (SIFT) have been considered for pattern recognition. These features are extracted from images according to manual predefined algorithms. In recent years, Machine Learning (ML) and Neural Networks (NNs) have been used for emotion recognition. In this report, a Convolutional Neural Network (CNN) is used to extract features from images to detect emotions. The Python Dlib toolkit is used to identify and extract 64 important landmarks on a face. A CNN model is trained with grayscale images from the FER 2013 dataset to classify expressions into five emotions namely happy, sad, neutral, fear and angry. To improve the accuracy and avoid overfitting of the model, batch normalization and dropout are used. The best value for each parameter is determined considering the training results. The results obtained show that CNN Model 1 is 80% accurate for four emotions (happy, sad, angry, fear) and 72% accurate for five emotions (happy, sad, angry, neutral, fear), while CNN Model 2 is 79% accurate for four emotions and 72% accurate for five emotions.