Voxel-based 3D occlusion-invariant face recognition using game theory and simulated annealing
- PDF / 2,324,048 Bytes
- 31 Pages / 439.37 x 666.142 pts Page_size
- 24 Downloads / 160 Views
Voxel-based 3D occlusion-invariant face recognition using game theory and simulated annealing Sahil Sharma 1
& Vijay Kumar
2
Received: 8 December 2019 / Revised: 13 June 2020 / Accepted: 9 July 2020 # Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract
A novel voxel-based occlusion-invariant 3D face recognition framework (V3DOFR) based on game theory and simulated annealing is proposed. In V3DOFR approach, 3D meshes are converted to voxel form of sizes 43, 83, and 163. After that, locality preserving projection-based embeddings are computed for removing the sparseness of voxels and generating consistent linear embedding per mesh with size 64 × 3, 128 × 3, and 256 × 3, respectively. The generator of triplets provides the triplets of sizes 64x3x3, 128x3x3, and 256x3x3. The simulated annealing is used to check the threshold value of adversarial triplet loss generated after ensembling losses of different grid sizes. The proposed framework is compared with four well-known methods using three face datasets, namely, Bosphorus, UMBDB, and KinectFaceDB. The performance evaluation has been done using four different cases of experimentations, viz. voxel based face recognition, occlusion invariant face recognition, landmarks based 3D face recognition, and 3D mesh based face recognition. Seven evaluation metrics are used to compare the proposed technique with other methods. The proposed method provides better accuracy and computation time over the other existing techniques in the majority of cases. Keywords 3Dmesh . Voxelization . Adversarial triplet loss . Generator . Discriminator . Simulated annealing
* Sahil Sharma [email protected] Vijay Kumar [email protected]
1
Computer Science and Engineering Department, Thapar Institute of Engineering and Technology, Patiala, India
2
Computer Science and Engineering Department, National Institute of Technology, Hamirpur, India
Multimedia Tools and Applications
1 Introduction 3D face recognition is widely used throughout the world due to the availability of easily collectable 3D data and capabilities of computation with the availability of highly economical graphical processing units (GPUs). However, acquiring 3D images are harder as compared to 2D scans. Therefore, the number of images is limited in public databases [25, 86, 90]. In [90], a high resolution spontaneous 3D dynamic facial expression database is presented. This work supports 3D spatiotemporal features exploration in subtle face expression. In [86], high resolution data acquisition is done using 3D dynamic imaging system setup. There are total of 101 number of subjects, six unique expressions, 606 number of 2D texture videos, 606 number of 3D model sequences, and approximately 60,600 3D models. In [25], 3D face recognition is improved using multi-instance enrollment representation. The experiments were performed on ND-2006 3D face dataset [57], that consists of 13,450 3D images. There are various techniques available in the literature for handling 3D mesh data, RGB-D image,
Data Loading...
