3D Try on Using Deep Neural Networks

Sai Medavarapu

Abstract: This paper proposes a virtual try on system where user can try the eye glasses on the 2D face in real time. The fundamental specialized test of this framework is the programmed 3D eyeglasses model remaking from the 2D glasses on a frontal human face. This paper initially proposes a convolutional neural network model to label the facial key points. Generated model is used for placing the glasses on the detected facial key points. The test results exhibit the efficiency on execution of this methodology and the business capability of proposed methodology. Code is made available at https://gitlab.tamucc.edu/smedavarapu1/ase-capstone.git.

Keywords: Facial landmark detection, Human Activity Recognition, Masked object detection, contour detection, 3D Try on