Multispectral Periocular Classification With Multimodal Compact Multi-Linear Pooling

AlGashaam, Faisal; Nguyen, Kien; Alkanhal, Mohamed; Chandran, Vinod; Boles, Wageeh; Banks, Jasmine

doi:10.1109/access.2017.2731118

Cited by 20 publications

(13 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The first three are basic multimodal integration models, which are early integration, late integration, and intermediate integration. The rest two are an intermediate integration model with compact multi-linear pooling [50] and an early integration model pretrained by a multimodal autoencoder [41]. To ensure fair comparison, we design the network models to have the same structure except for the part for integration.…”

Section: Network Modelsmentioning

confidence: 99%

“…We employ the compact multi-linear pooling method [39,50]. As shown in Figures 3 (d) and 4 (d), the structures of the models are essentially the same to those of the intermediate integration models, except for the integration part.…”

Section: Compact Multi-linear Poolingmentioning

confidence: 99%

“…As shown in Figures 3 (d) and 4 (d), the structures of the models are essentially the same to those of the intermediate integration models, except for the integration part. By following the previous work [39,50], each model first calculates the count sketch [38] for each modality feature (Ψ in Figures 3 (d) and 4 (d)), an element-wise multiplication is performed to the Fourier-transformed outputs of the count sketches, and the inverse Fourier transform is applied to the multiplied result. In addition, a L2 normalization is applied to each modality before calculating its count sketch, as in the previous work [39,53].…”

Section: Compact Multi-linear Poolingmentioning

confidence: 99%

“…Structures of the employed network models for the OPPORTUNITY dataset. (a) Early integration (b) Late integration (c) Intermediate integration (d)Compact multi-linear pooling[50] (e) Multimodal autoencoder[41] (f) EmbraceNet…”

mentioning

confidence: 99%

See 3 more Smart Citations

EmbraceNet: A robust deep learning architecture for multimodal classification

2019

View full text Add to dashboard Cite

Classification using multimodal data arises in many machine learning applications. It is crucial not only to model cross-modal relationship effectively but also to ensure robustness against loss of part of data or modalities. In this paper, we propose a novel deep learning-based multimodal fusion architecture for classification tasks, which guarantees compatibility with any kind of learning models, deals with cross-modal information carefully, and prevents performance degradation due to partial absence of data. We employ two datasets for multimodal classification tasks, build models based on our architecture and other state-of-the-art models, and analyze their performance on various situations. The results show that our architecture outperforms the other multimodal fusion architectures when some parts of data are not available.

show abstract

Section: Network Modelsmentioning

confidence: 99%

Section: Compact Multi-linear Poolingmentioning

confidence: 99%

Section: Compact Multi-linear Poolingmentioning

confidence: 99%

mentioning

confidence: 99%

See 2 more Smart Citations

EmbraceNet: A robust deep learning architecture for multimodal classification

2019

View full text Add to dashboard Cite

show abstract

“…In recent years, it has gained more attention worldwide. Several biometric traits, including face, gait, iris, key-stroke, fingerprint, and palmprint have been widely studied and developed depending on the application domain most suitable for them [ 2 , 3 , 4 ]. Compared to other biometrics, palmprint has a low distortion, strong stability, and high uniqueness [ 5 ].…”

Section: Introductionmentioning

confidence: 99%

An Effective Palmprint Recognition Approach for Visible and Multispectral Sensor Images

Gumaei

Sammouda

Al-Salman

et al. 2018

Sensors

View full text Add to dashboard Cite

Among several palmprint feature extraction methods the HOG-based method is attractive and performs well against changes in illumination and shadowing of palmprint images. However, it still lacks the robustness to extract the palmprint features at different rotation angles. To solve this problem, this paper presents a hybrid feature extraction method, named HOG-SGF that combines the histogram of oriented gradients (HOG) with a steerable Gaussian filter (SGF) to develop an effective palmprint recognition approach. The approach starts by processing all palmprint images by David Zhang’s method to segment only the region of interests. Next, we extracted palmprint features based on the hybrid HOG-SGF feature extraction method. Then, an optimized auto-encoder (AE) was utilized to reduce the dimensionality of the extracted features. Finally, a fast and robust regularized extreme learning machine (RELM) was applied for the classification task. In the evaluation phase of the proposed approach, a number of experiments were conducted on three publicly available palmprint databases, namely MS-PolyU of multispectral palmprint images and CASIA and Tongji of contactless palmprint images. Experimentally, the results reveal that the proposed approach outperforms the existing state-of-the-art approaches even when a small number of training samples are used.

show abstract

Deep Spectral Biometrics: Overview and Open Issues

Munir¹,

Khan²

2020

Deep Biometrics

View full text Add to dashboard Cite

Multispectral Periocular Classification With Multimodal Compact Multi-Linear Pooling

Cited by 20 publications

References 34 publications

EmbraceNet: A robust deep learning architecture for multimodal classification

EmbraceNet: A robust deep learning architecture for multimodal classification

An Effective Palmprint Recognition Approach for Visible and Multispectral Sensor Images

Deep Spectral Biometrics: Overview and Open Issues

Contact Info

Product

Resources

About