I would like to express my gratitude to my supervisor, Mr Rechard Lee, for his guidance throughout the project. Furthermore, I would like to thank my lecturers, my family and friends that had also supported me during this project
Illumination Test
The objectives of this project are:
1. To develop a markerless-based system that can detect the face and to extract the facial features.
2. To develop a system that can overlay the 3D rendered elements.
In this project, an augmented reality face filter was built.
The proposed system results in high accuracy rate face detection which is 98.1%. despite having different lighting intensity, the 3D element was still rendered from the system.
To develop a marker-less based system that can detect the facial features and overlay 3D elements
In the recent years, there has been an influx of applications that use face filters incorporating augmented reality technology. In this project, a markerless-based AR system will be developed, where 3D rendered elements will be projected onto the face of the subject. As there are problems in detecting faces as well as the facial features to be designated as markers for the 3D element. the 3D elements must be able to be overlaid onto
the face in various lighting conditions and orientations
SW40106 PROJEK SAINTIFIK II
MATHEMATICS WITH COMPUTER GRAPHICS PROGRAMME
NUR AIDA AQILA BINTI MOHAMED AZAHARI & RECHARD LEE
FAKULTI SAINS DAN SUMBER ALAM, UNIVERSITI MALAYSIA SABAH , 88400, KOTA KINABALU, SABAH [email protected] & [email protected]
AUGMENTED REALITY FACE FILTER USING ARCORE
In this era of globalization, augmented reality (AR) is widely used in varies fields such as in gaming, medical andeducation. AR is a combination of real and virtual environments. In this project, an AR face filter using ARCore is proposed. Performance and illuminance variation tests are carried
out to further analyze and evalute the system.
ABSTRACT
INTRODUCTION
METHODOLOGY AIM
CONCLUSION
ACKNOWLEDGEMENTS OBJECTIVE
RESULTS AND DISCUSSION
Testing the system on various faces (human &
non-human) achieves an accuracy of 98.1%
Distance Test
Due to limitation of ARCore, faces are only able to be detected up to 1m away. Faces are detected accurately when below 1m away.
Occlusion Test Accuracy Test
Sample Output
System is able to detect face best when face is 75% or more visible.
When testing the system in 10, 30, 200, and 300 Lux, the system was able to detect the face and display 3D element accurately.