SpPCANet: a simple deep learning-based feature extraction approach for 3D face recognition
- PDF / 2,035,367 Bytes
- 24 Pages / 439.37 x 666.142 pts Page_size
- 109 Downloads / 206 Views
SpPCANet: a simple deep learning-based feature extraction approach for 3D face recognition Koushik Dutta 1
1
& Debotosh Bhattacharjee & Mita Nasipuri
1
Received: 10 August 2019 / Revised: 30 June 2020 / Accepted: 6 August 2020 # Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract A Sparse Principal Component Analysis Network (SpPCANet) based feature extraction is proposed here for 3D face recognition. The network consists of three basic components: (1) Multistage sparse principal component analysis filters, (2) Binary hashing, and (3) Block-wise histogram computation. Here, the sparse principal component analysis is used to learn multistage filter banks at the convolution stage, which is followed by binary hashing for indexing and block-wise histogram for pooling. Finally, a linear support vector machine (SVM) is used for classifying the features extracted by SpPCANet. The proposed network SpPCANet is a lightweight deep learning network. Three well-known 3D face databases, namely, Frav3D, Bosphorus3D, and Casia3D, are used for validating the proposed system. This proposed network has been extensively studied by varying different parameters, such as the number of filters at the convolution layer and the size of filters at the convolution layer and size of non-overlapping blocks at the pooling layer. Handling all types of variation of faces available in Frav3D, Bosphorus3D, and Casia3D databases, the system has acquired 96.93%, 98.54%, and 88.80% recognition rates, respectively. Keywords 3D face image . Sparse principal component analysis filter . Binary hashing . Blockwise histogram . Lightweight deep network
* Koushik Dutta [email protected] Debotosh Bhattacharjee [email protected] Mita Nasipuri [email protected]
1
Computer Science and Engineering, Jadavpur University, 188, Raja S. C. Maulik Road, Kolkata 700032, India
Multimedia Tools and Applications
1 Introduction Convolutional neural network (CNN) or Convnet [34, 42] is currently the most popular tool in computer vision. Various CNN archietecture have been designed for image segmentation, feature extraction, classification, and object recognition. Most of the cases, CNN achieves state-of-the-art results on various image databases. In any classification/recognition problem, researchers mainly focus on extracting innovative features, which is one of the part in classification. However, finding innovative features and selecting a more informative subset of them for reducing the complexity of the system are the limitations of hand-crafted features. The learning-based feature extraction approach of CNN has overcome these limitations. In CNN, the key challenges are to design a proper network architecture and choosing the right configuration and parameters such as the number of layers, filter size, choice of pooling function, etc. There exist various convolution network archietectures like LeNet-5 [23], AlexNet [22], GoogleNet [34], FaceNet [31], etc.; which are used in computer vision. CNN is recently used in various
Data Loading...
