FV-ViT: Vision Transformer for Finger Vein Recognition

Li, Xiaoye; Zhang, Binbin

doi:10.1109/access.2023.3297212

Cited by 10 publications

(3 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…From the table, it is evident that our lightweight vein recognition algorithm, which combines CNN and Transformer, outperforms algorithms using Transformer as the primary feature extractor. In comparison with Li [ 44 ], our approach achieves a substantial increase in recognition accuracy. Specifically, on the SDU and USM datasets, our algorithm improves recognition accuracy by 7.29% and 0.27%, respectively, while also reducing the EER by 0.55% and 0.096%.…”

Section: Experiments and Resultsmentioning

confidence: 92%

“…The first two studies utilize Convolutional Neural Networks (CNNs), achieving high recognition accuracy. Huang et al [ 32 ] employed a Transformer as the primary feature extractor, while Li et al [ 44 ] used a Vision Transformer for feature extraction. From the table, it is evident that our lightweight vein recognition algorithm, which combines CNN and Transformer, outperforms algorithms using Transformer as the primary feature extractor.…”

Section: Experiments and Resultsmentioning

confidence: 99%

See 1 more Smart Citation

FV-MViT: Mobile Vision Transformer for Finger Vein Recognition

Li,

Feng,

Cai

et al. 2024

Sensors

View full text Add to dashboard Cite

In addressing challenges related to high parameter counts and limited training samples for finger vein recognition, we present the FV-MViT model. It serves as a lightweight deep learning solution, emphasizing high accuracy, portable design, and low latency. The FV-MViT introduces two key components. The Mul-MV2 Block utilizes a dual-path inverted residual connection structure for multi-scale convolutions, extracting additional local features. Simultaneously, the Enhanced MobileViT Block eliminates the large-scale convolution block at the beginning of the original MobileViT Block. It converts the Transformer’s self-attention into separable self-attention with linear complexity, optimizing the back end of the original MobileViT Block with depth-wise separable convolutions. This aims to extract global features and effectively reduce parameter counts and feature extraction times. Additionally, we introduce a soft target center cross-entropy loss function to enhance generalization and increase accuracy. Experimental results indicate that the FV-MViT achieves a recognition accuracy of 99.53% and 100.00% on the Shandong University (SDU) and Universiti Teknologi Malaysia (USM) datasets, with equal error rates of 0.47% and 0.02%, respectively. The model has a parameter count of 5.26 million and exhibits a latency of 10.00 milliseconds from the sample input to the recognition output. Comparison with state-of-the-art (SOTA) methods reveals competitive performance for FV-MViT.

show abstract

Section: Experiments and Resultsmentioning

confidence: 92%

Section: Experiments and Resultsmentioning

confidence: 99%

FV-MViT: Mobile Vision Transformer for Finger Vein Recognition

Li,

Feng,

Cai

et al. 2024

Sensors

View full text Add to dashboard Cite

show abstract

“…Excellent recognition performance was achieved after adding strict regularization. Devkota et al [35] integrated the DenseNet model, squeeze excitation (SE), and hybrid pooling (HP). Then, a series of preprocessing methods were used to obtain vein patterns.…”

Section: Finger Vein Image Enhancement Based On Deep Featurementioning

confidence: 99%

A Degraded Finger Vein Image Recovery and Enhancement Algorithm Based on Atmospheric Scattering Theory

Feng,

Mao

et al. 2024

Sensors

View full text Add to dashboard Cite

With the development of biometric identification technology, finger vein identification has received more and more widespread attention for its security, efficiency, and stability. However, because of the performance of the current standard finger vein image acquisition device and the complex internal organization of the finger, the acquired images are often heavily degraded and have lost their texture characteristics. This makes the topology of the finger veins inconspicuous or even difficult to distinguish, greatly affecting the identification accuracy. Therefore, this paper proposes a finger vein image recovery and enhancement algorithm using atmospheric scattering theory. Firstly, to normalize the local over-bright and over-dark regions of finger vein images within a certain threshold, the Gamma transform method is improved in this paper to correct and measure the gray value of a given image. Then, we reconstruct the image based on atmospheric scattering theory and design a pixel mutation filter to segment the venous and non-venous contact zones. Finally, the degraded finger vein images are recovered and enhanced by global image gray value normalization. Experiments on SDUMLA-HMT and ZJ-UVM datasets show that our proposed method effectively achieves the recovery and enhancement of degraded finger vein images. The image restoration and enhancement algorithm proposed in this paper performs well in finger vein recognition using traditional methods, machine learning, and deep learning. The recognition accuracy of the processed image is improved by more than 10% compared to the original image.

show abstract

Finger Vein Recognition Based on ResNet With Self-Attention

Zhang,

Chen,

Zhang

et al. 2024

IEEE Access

View full text Add to dashboard Cite

To solve the problem of low accuracy and high computational resource consumption in finger vein recognition, a finger vein recognition model based on ResNet with self-attention (FV-RSA) is proposed. This model combines global focusing ability of self-attention mechanism and local feature extraction ability of CNN, which improves recognition accuracy. To reduce the number of parameters and floating-point operations, self-attention and convolution share linear projections by pointwise convolution. Self-attention and CNN are fused in the convolution and self-attention (CASA) block connected by skip connection to avoid gradient vanishing or gradient exploding. During the training phase, we use a variable learning rate with cosine annealing to avoid falling into local optimum. Experiments show that the method works well on the public database, which can not only improve the accuracy, but also reduce the number of parameters and computational complexity. INDEX TERMS finger vein recognition, deep learning, ResNet, self-attention, variable learning rate.

show abstract

FV-ViT: Vision Transformer for Finger Vein Recognition

Cited by 10 publications

References 35 publications

FV-MViT: Mobile Vision Transformer for Finger Vein Recognition

FV-MViT: Mobile Vision Transformer for Finger Vein Recognition

A Degraded Finger Vein Image Recovery and Enhancement Algorithm Based on Atmospheric Scattering Theory

Finger Vein Recognition Based on ResNet With Self-Attention

Contact Info

Product

Resources

About