Search

Finger vein recognition is an advanced biometric recognition technology that offers high precision and high security. It recognizes or authenticates individuals using irradiating vein texture images collected from fingers with near-infrared light. In this paper, we propose a new finger vein recognition model (MMRAN) based on a multiscale and multistage residual attention network. First, to fully adapt to the low-resolution, grayscale pixels, and linear patterns of finger vein images, we designed an architecture that combines a fusion residual attention block (FRAB) and a multistage residual attention connection (MRAC). The FRAB contains two distinct subpaths: the main vein path (MVP) and the guided attention path (GAP). The MVP extracts finger vein features at multiple scales by using a multibranch residual structure, while the GAP uses an hourglass network to generate weight maps to guide the setting of eigenvalues at corresponding locations in the main vein feature map. MRAC integrates venous features extracted at different learning stages through the above two pathways. The proposed multiscale and multistage extraction model is effective at extracting various types of digital vein features, including those whose shapes change in width, direction, curvature, and so on. We combine the various dimensions of vein features through multistage learning to further improve the model performance for extracting high-level abstract features. To evaluate the performance of our proposed model, we conducted a large number of experiments on five publicly available finger vein datasets. The experimental results show that the proposed model not only achieves a recognition accuracy above 98%, which is an improvement compared to the current state-of-the-art methods, but it can also be implemented with fewer parameters, which improves training and inference. [ABSTRACT FROM AUTHOR]

With the development of deep learning technology, many deep learning methods have been applied to font recognition and generation. However, few studies focus on font inpainting problems. This paper is dedicated to repairing damaged fonts based on style to repair damaged fonts in a better way. In this paper, we propose a CGAN (Conditional Generative Adversarial Nets)-based font repair method. This paper uses the content accuracy and style similarity of the repaired image as an evaluation index to evaluate the accuracy of the restored style font. The font content proposed by the paper based on CGAN network repair style is similar with the correct content. [ABSTRACT FROM AUTHOR]

Deep-sea organism automatic tracking has rarely been studied because of a lack of training data. However, it is extremely important for underwater robots to recognize and to predict the behavior of organisms. In this paper, we first develop a method for underwater real-time recognition and tracking of multi-objects, which we call "You Only Look Once: YOLO". This method provides us with a very fast and accurate tracker. At first, we remove the haze, which is caused by the turbidity of the water from a captured image. After that, we apply YOLO to allow recognition and tracking of marine organisms, which include shrimp, squid, crab and shark. The experiments demonstrate that our developed system shows satisfactory performance. [ABSTRACT FROM AUTHOR]

Convolutional networks are currently the most popular computer vision methods for a wide variety of applications in multimedia research fields. Most recent methods have focused on solving problems with natural images and usually use a training database, such as Imagenet or Openimage, to detect the characteristics of the objects. However, in practical applications, training samples are difficult to acquire. In this study, we develop a powerful approach that can accurately learn marine organisms. The proposed filtering deep convolutional network (FDCNet) classifies deep-sea objects better than state-of-the-art classification methods, such as AlexNet, GoogLeNet, ResNet50, and ResNet101. The classification accuracy of the proposed FDCNet method is 1.8%, 2.9%, 2.0%, and 1.0% better than AlexNet, GooLeNet, ResNet50, and ResNet101, respectively. In addition, we have built the first marine organism database, Kyutech10K, with seven categories (i.e., shrimp, squid, crab, shark, sea urchin, manganese, and sand). [ABSTRACT FROM AUTHOR]

The original version of this article unfortunately contained a mistake in the Affiliation section. [ABSTRACT FROM AUTHOR]