Realtids ansiktsigenkänning på inbäddat system tillämpat på set-top-box
Year: 2025
Type: dissertation
Abstract: This thesis explores state-of-the-art of face recognition models applied to an embedded platform, with a focus on optimizing performance for real-time ap- plications. Face recognition technology, a crucial aspect of modern secu- rity and biometric systems, typically demands significant computational re- sources. However, the constraints of embedded systems, such as limited pro- cessing power, memory, and energy efficiency, limits the use of deep learning models. This research addresses these challenges by constructing a lightweight, efficient face recognition system for embedded hardware. The proposed sys- tem leverages optimization techniques to execute machine learning models on low-power processors. Through comprehensive testing, including accuracy, processing speed, and resource utilization, the system demonstrates a balance between performance and efficiency, making it suitable for various applica- tions like Netflix on a set-top box, but it can be extended to other edged devices. The results indicate that the implemented face recognition system can operate effectively within the resource constraints of embedded systems, paving the way for real-time, on-device biometric applications.
Author: Xu, Roxanne
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W6996128678