A Novel Approach in detecting pose orientation of a 3D face required for face

TL;DR

This paper introduces a new method for detecting the pose orientation of 3D faces, identifying the axis and rotation angles with reasonable accuracy, based on experiments on the FRAV3D database.

Contribution

The paper presents a novel algorithm for 3D face pose detection that works with rotation angles up to 40 degrees, demonstrating effectiveness on a large dataset.

Findings

67% correct pose detection on 848 images

Effective for rotations up to 40 degrees

Validated on FRAV3D database

Abstract

In this paper we present a novel approach that takes as input a 3D image and gives as output its pose i.e. it tells whether the face is oriented with respect the X, Y or Z axes with angles of rotation up to 40 degree. All the experiments have been performed on the FRAV3D Database. After applying the proposed algorithm to the 3D facial surface we have obtained i.e. on 848 3D face images our method detected the pose correctly for 566 face images,thus giving an approximately 67 % of correct pose detection.