Is this project an undergraduate, graduate, or faculty project?
Undergraduate
Project Type
group
Campus
Daytona Beach
Authors' Class Standing
Andrew Murphy, Senior Giovanni C. DeCapua, Senior Anthony M. Cafiso, Senior Max E. Raabe, Senior Kaden E. Van Leuven, Senior
Lead Presenter's Name
Andrew Murphy
Lead Presenter's College
DB College of Engineering
Faculty Mentor Name
Dr. Sirani Mututhanthrige Perera
Abstract
WFOV lenses are becoming popular in facial recognition due to the fact that they enhance subject coverage and improve the chances of detecting target faces. However, wide-angle optics introduce nonlinear distortion around the image periphery, which degrades the performance of recognition pipelines. In this poster presentation, we use WFOV lens captures to analyze facial recognition using classical low-complexity algorithms based on the discrete Fourier transform (DFT), discrete cosine transform (DCT), principal component analysis (PCA), and data-driven learning with convolutional neural networks. Finally, we present computational efficiency, compression, accuracy, and precision of recognizing distorted images with qualitative and quantitative measures.
Did this research project receive funding support (Spark, SURF, Research Abroad, Student Internal Grants, Collaborative, Climbing, or Ignite Grants) from the Office of Undergraduate Research?
No
Included in
A Comparative Classical and Data-driven Facial Analysis of Wide-Field-of-View Lens Captures
WFOV lenses are becoming popular in facial recognition due to the fact that they enhance subject coverage and improve the chances of detecting target faces. However, wide-angle optics introduce nonlinear distortion around the image periphery, which degrades the performance of recognition pipelines. In this poster presentation, we use WFOV lens captures to analyze facial recognition using classical low-complexity algorithms based on the discrete Fourier transform (DFT), discrete cosine transform (DCT), principal component analysis (PCA), and data-driven learning with convolutional neural networks. Finally, we present computational efficiency, compression, accuracy, and precision of recognizing distorted images with qualitative and quantitative measures.