| Author: | Li, Kunluo |
| Title: | Facial expression recognition using attention mechanisms |
| Degree: | M.Sc. |
| Year: | 2024 |
| Department: | Department of Electrical and Electronic Engineering |
| Pages: | ix, 36 pages : color illustrations |
| Language: | English |
| Abstract: | Facial expressions are important in human communication. They show emotions and intentions without words. This paper presents a new Facial Expression Recognition (FER) model. The overall architecture is based on combining the architecture of MobileNetV3 and TinyViT; a model also employing multi-head attention mechanisms. Our approach performs local-global features on facial expressions, boosting performance and strength in detecting a true emotion. MobileNetV3 extracts detailed local features from key facial areas. TinyViT, a transformer-based network, models global dependencies. It recognizes subtle expressions with high precision. The models are integrated using an 8-head multi-head attention mechanism. This allows the system to focus on the most informative parts of the face. It improves feature fusion and overall performance. We performed extensive experiments on three benchmark datasets: RAF-DB, FER2013, and AffectNet. Our hybrid model achieved an accuracy of 82.11%. This shows its ability to handle diverse real-world challenges. These challenges include varying lighting conditions, occlusions, and demographic differences. The model has difficulty accurately recognizing subtle emotions like 'fear' and 'disgust'. These results show that there is potential for further improvements. Future work can address class imbalance and optimize computational efficiency. |
| Rights: | All rights reserved |
| Access: | restricted access |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| 8309.pdf | For All Users (off-campus access for PolyU Staff & Students only) | 1.68 MB | Adobe PDF | View/Open |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/13901