|
Title:
|
REAL-TIME HEAD POSE ESTIMATION WITH SVM MODEL FOR FRONTAL FACE CLASSIFICATION |
|
Author(s):
|
Loubrys L. Rojas Reinoso, Fernando L. Gutiérrez López, José C. Gutiérrez, Graça Bressan and Wilson Vicente Ruggiero
|
|
ISBN:
|
978-989-8533-99-9 |
|
Editors:
|
Piet Kommers, Adriana Backx Noronha Viana, Theodora Issa,
Pedro Isaías and Tomayess Issa |
|
Year:
|
2020 |
|
Edition:
|
Single |
|
Keywords:
|
Degree of Freedom of Human Head (DOF), Head Pose Estimation (HPE), Support Vector Machine (SVM), Frontal Face
Classification, Machine Learning, Computer Vi |
|
Type:
|
Short |
|
First Page:
|
63 |
|
Last Page:
|
67 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
click to dowload 
|
|
Paper Abstract:
|
Head pose estimation (HPE) has been widely studied in the last years due to its many applications in face analysis systems.
The use of such systems ranges from the analysis of focus of attention, social interactions or the use in mobile applications
in the realization of the currently popular facial animations and / or in face recognition process, where the frontal faces are
especially important. Many approaches were proposed focusing mainly in Random Forests and Convolutional Neural
Networks (CNN). In this paper, a framework for estimation of the head pose was proposed computing the degrees of
freedom (DOF) of the human head using 2D images data only. The framework implements some computer vision
algorithms available in publicly machine learning libraries such as OpenCV and Dlib, which allows easy application and
re-implementation. In addition, a Support Vector Machine (SVM) model with Radial Basis Function (RBF) kernel was
developed for frontal face classification. Experiments conducted on 2D image datasets in constrained environment show
that the approach is capable of real-time performance. Were designed three protocols of experiments with two databases
for testing the SVM model. Values of 100% and 98% for precision and recall, respectively, were achieved classifying
frontal faces. Significant results were obtained measuring yaw rotation with 4.24 of mean absolute error for frontal face. |
|
|
|
|
|
|
