Real-time matting is a challenging research in deep learning. Conventional CNN (Convolutional Neural Networks) approaches are easy to misjudge the foreground and background semantic and have blurry matting edges, which result from CNN’s limited concentration on global context due to receptive field. We propose a real-time matting approach called RMViT (Real-time matting with Vision Transformer) with Transformer structure, attention and content-aware guidance to solve issues above. The semantic accuracy improves a lot due to the establishment of global context and long-range pixel information. The experiments show our approach exceeds a 30% reduction in error metrics compared with existing real-time matting approaches.

The combination of multiple-input multiple-output (MIMO) technology and sparse code multiple access (SCMA) can significantly enhance the spectral efficiency of future wireless communication networks. However, the receiver design for downlink MIMO-SCMA systems faces challenges in developing multi-user detection (MUD) schemes that achieve both low latency and low bit error rate (BER). The separated detection scheme in the MIMO-SCMA system involves performing MIMO detection first to obtain estimated signals, followed by SCMA decoding. We propose an enhanced separated detection scheme based on lightweight graph neural networks (GNNs). In this scheme, we raise the concept of coordinate point relay and full-category training, which allow for the substitution of the conventional message passing algorithm (MPA) in SCMA decoding with image classification techniques based on deep learning (DL). The features of the images used for training encompass crucial information such as the amplitude and phase of estimated signals, as well as channel characteristics they have encountered. Furthermore, various types of images demonstrate distinct directional trends, contributing additional features that enhance the precision of classification by GNNs. Simulation results demonstrate that the enhanced separated detection scheme outperforms existing separated and joint detection schemes in terms of computational complexity, while having a better BER performance than the joint detection schemes at high Eb/N0 (energy per bit to noise power spectral density ratio) values.

Code search is a task to retrieve the most relevant code given a natural language query. Several recent studies proposed deep learning based methods use multi-encoder model to parse code into multi-field to represent code. These methods enhance the performance of the model by distinguish between similar codes and utilizing a relation matrix to bridge the code and query. However, these models require more computational resources and parameters than single-encoder models. Furthermore, utilizing the relation matrix that solely relies on max-pooling disregards the delivery of word alignment information. To alleviate these problems, we propose a combined alignment model for code search. We concatenate the multi-code fields into one sequence to represent code and use one encoding model to encode code features. Moreover, we transform the relation matrix using trainable vectors to avoid information losses. Then, we combine intra-modal and cross-modal attention to assign the salient words while matching the corresponding code and query. Finally, we apply the attention weight to code/query embedding and compute the cosine similarity. To evaluate the performance of our model, we compare our model with six previous models on two popular datasets. The results show that our model achieves 0.614 and 0.687 Top@1 performance, outperforming the best comparison models by 12.2% and 9.3%, respectively.

Depression is a prevalent mental disorder affecting a significant portion of the global population, leading to considerable disability and contributing to the overall burden of disease. Consequently, designing efficient and robust automated methods for depression detection has become imperative. Recently, deep learning methods, especially multimodal fusion methods, have been increasingly used in computer-aided depression detection. Importantly, individuals with depression and those without respond differently to various emotional stimuli, providing valuable information for detecting depression. Building on these observations, we propose an intra- and inter-emotional stimulus transformer-based fusion model to effectively extract depression-related features. The intra-emotional stimulus fusion framework aims to prioritize different modalities, capitalizing on their diversity and complementarity for depression detection. The inter-emotional stimulus model maps each emotional stimulus onto both invariant and specific subspaces using individual invariant and specific encoders. The emotional stimulus-invariant subspace facilitates efficient information sharing and integration across different emotional stimulus categories, while the emotional stimulus specific subspace seeks to enhance diversity and capture the distinct characteristics of individual emotional stimulus categories. Our proposed intra- and inter-emotional stimulus fusion model effectively integrates multimodal data under various emotional stimulus categories, providing a comprehensive representation that allows accurate task predictions in the context of depression detection. We evaluate the proposed model on the Chinese Soochow University students dataset, and the results outperform state-of-the-art models in terms of concordance correlation coefficient (CCC), root mean squared error (RMSE) and accuracy.

Acoustic scene classification (ASC) is a fundamental domain within the realm of artificial intelligence classification tasks. ASC-based tasks commonly employ models based on convolutional neural networks (CNNs) that utilize log-Mel spectrograms as input for gathering acoustic features. In this paper, we designed a CNN-based multi-scale pooling (MSP) strategy for ASC. The log-Mel spectrograms are utilized as the input to CNN, which is partitioned into four frequency axis segments. Furthermore, we devised four CNN channels to acquire inputs from distinct frequency ranges. The high-level features extracted from outputs in various frequency bands are integrated through frequency pyramid average pooling layers at multiple levels. Subsequently, a softmax classifier is employed to classify different scenes. Our study demonstrates that the implementation of our designed model leads to a significant enhancement in the model's performance, as evidenced by the testing of two acoustic datasets.

Noisy labels in training data can significantly harm the performance of deep neural networks (DNNs). Recent research on learning with noisy labels uses a property of DNNs called the memorization effect to divide the training data into a set of data with reliable labels and a set of data with unreliable labels. Methods introducing semi-supervised learning strategies discard the unreliable labels and assign pseudo-labels generated from the confident predictions of the model. So far, this semi-supervised strategy has yielded the best results in this field. However, we observe that even when models are trained on balanced data, the distribution of the pseudo-labels can still exhibit an imbalance that is driven by data similarity. Additionally, a data bias is seen that originates from the division of the training data using the semi-supervised method. If we address both types of bias that arise from pseudo-labels, we can avoid the decrease in generalization performance caused by biased noisy pseudo-labels. We propose a learning method with noisy labels that introduces unbiased pseudo-labeling based on causal inference. The proposed method achieves significant accuracy gains in experiments at high noise rates on the standard benchmarks CIFAR-10 and CIFAR-100.

In the beyond 5G and 6G networks, the number of connected devices and their types will greatly increase including not only user devices such as smartphones but also the Internet of Things (IoT). Moreover, Non-terrestrial networks (NTN) introduce dynamic changes in the types of connected devices as base stations or access points are moving objects. Therefore, continuous network capacity design is required to fulfill the network requirements of each device. However, continuous optimization of network capacity design for each device within a short time span becomes difficult because of the heavy calculation amount. We introduce device types as groups of devices whose traffic characteristics resemble and optimize network capacity per device type for efficient network capacity design. This paper proposes a method to classify device types by analyzing only encrypted traffic behavior without using payload and packets of specific protocols. In the first stage, general device types, such as IoT and non-IoT, are classified by analyzing packet header statistics using machine learning. Then, in the second stage, connected devices classified as IoT in the first stage are classified into IoT device types, by analyzing a time series of traffic behavior using deep learning. We demonstrate that the proposed method classifies device types by analyzing traffic datasets and outperforms the existing IoT-only device classification methods in terms of the number of types and the accuracy. In addition, the proposed model performs comparable as a state-of-the-art model of traffic classification, ResNet 1D model. The proposed method is suitable to grasp device types in terms of traffic characteristics toward efficient network capacity design in networks where massive devices for various services are connected and the connected devices continuously change.

The fingerprint verification system is widely used in mobile devices because of fingerprint's distinctive features and ease of capture. Typically, mobile devices utilize small sensors, which have limited area, to capture fingerprint. Meanwhile, conventional fingerprint feature extraction methods need detailed fingerprint information, which is unsuitable for those small sensors. This paper proposes a novel fingerprint verification method for small area sensors based on deep learning. A systematic method combines deep convolutional neural network (DCNN) in a Siamese network for feature extraction and XGBoost for fingerprint similarity training. In addition, a padding technique also introduced to avoid wraparound error problem. Experimental results show that the method achieves an improved accuracy of 66.6% and 22.6% in the FingerPassDB7 and FVC2006DB1B dataset, respectively, compared to the existing methods.

It has become increasingly important for industries to promote digital transformation by utilizing 5G and industrial internet of things (IIoT) to improve productivity. To protect IIoT application performance (work speed, productivity, etc.), it is often necessary to satisfy quality of service (QoS) requirements precisely. For this purpose, there is an increasing need to automatically identify the root causes of radio-quality deterioration in order to take prompt measures when the QoS deteriorates. In this paper, a method for identifying the root cause of 5G radio-quality deterioration is proposed that uses machine learning. This Random Forest based method detects the root cause, such as distance attenuation, shielding, fading, or their combination, by analyzing the coefficients of a quadratic polynomial approximation in addition to the mean values of time-series data of radio quality indicators. The detection accuracy of the proposed method was evaluated in a simulation using the MATLAB 5G Toolbox. The detection accuracy of the proposed method was found to be 98.30% when any of the root causes occurs independently, and 83.13% when the multiple root causes occur simultaneously. The proposed method was compared with deep-learning methods, including bidirectional long short-term memory (bidirectional-LSTM) or one-dimensional convolutional neural network (1D-CNN), that directly analyze the time-series data of the radio quality, and the proposed method was found to be more accurate than those methods.

In this paper, we apply two methods in machine learning, dropout and semi-supervised learning, to a recently proposed method called CSQ-SDL which uses deep neural networks for evaluating shift quality from time-series measurement data. When developing a new Automatic Transmission (AT), calibration takes place where many parameters of the AT are adjusted to realize pleasant driving experience in all situations that occur on all roads around the world. Calibration requires an expert to visually assess the shift quality from the time-series measurement data of the experiments each time the parameters are changed, which is iterative and time-consuming. The CSQ-SDL was developed to shorten time consumed by the visual assessment, and its effectiveness depends on acquiring a sufficient number of data points. In practice, however, data amounts are often insufficient. The methods proposed here can handle such cases. For the cases wherein only a small number of labeled data points is available, we propose a method that uses dropout. For those cases wherein the number of labeled data points is small but the number of unlabeled data is sufficient, we propose a method that uses semi-supervised learning. Experiments show that while the former gives moderate improvement, the latter offers a significant performance improvement.

We assess how well the recently created MRI reconstruction technique, Multi-Resolution Convolutional Neural Network (MRCNN), performs in the core medical vision field (classification). The primary goal of MRCNN is to identify the best k-space undersampling patterns to accelerate the MRI. In this study, we use the Figshare brain tumor dataset for MRI classification with 3064 T1-weighted contrast-enhanced MRI (CE-MRI) over three categories: meningioma, glioma, and pituitary tumors. We apply MRCNN to the dataset, which is a method to reconstruct high-quality images from under-sampled k-space signals. Next, we employ the pre-trained VGG16 model, which is a Deep Neural Network (DNN) based image classifier to the MRCNN restored MRIs to classify the brain tumors. Our experiments showed that in the case of MRCNN restored data, the proposed brain tumor classifier achieved 92.79% classification accuracy for a 10% sampling rate, which is slightly higher than that of SRCNN, MoDL, and Zero-filling methods have 91.89%, 91.89%, and 90.98% respectively. Note that our classifier was trained using the dataset consisting of the images with full sampling and their labels, which can be regarded as a model of the usual human diagnostician. Hence our results would suggest MRCNN is useful for human diagnosis. In conclusion, MRCNN significantly enhances the accuracy of the brain tumor classification system based on the tumor location using under-sampled k-space signals.

Driver fatigue detection has become crucial in vehicle safety technology. Achieving high accuracy and real-time performance in detecting driver fatigue is paramount. In this paper, we propose a novel driver fatigue detection algorithm based on dynamic tracking of Facial Eyes and Yawning using YOLOv7, named FEY-YOLOv7. The Coordinate Attention module is inserted into YOLOv7 to enhance its dynamic tracking accuracy by focusing on coordinate information. Additionally, a small target detection head is incorporated into the network architecture to promote the feature extraction ability of small facial targets such as eyes and mouth. In terms of compution, the YOLOv7 network architecture is significantly simplified to achieve high detection speed. Using the proposed PERYAWN algorithm, driver status is labeled and detected by four classes: open_eye, closed_eye, open_mouth, and closed_mouth. Furthermore, the Guided Image Filtering algorithm is employed to enhance image details. The proposed FEY-YOLOv7 is trained and validated on RGB-infrared datasets. The results show that FEY-YOLOv7 has achieved mAP of 0.983 and FPS of 101. This indicates that FEY-YOLOv7 is superior to state-of-the-art methods in accuracy and speed, providing an effective and practical solution for image-based driver fatigue detection.

The diffusion model has achieved success in generating and editing high-quality images because of its ability to produce fine details. Its superior generation ability has the potential to facilitate more detailed segmentation. This study presents a novel approach to segmentation tasks using an inverse heat dissipation model, a kind of diffusion-based models. The proposed method involves generating a mask that gradually shrinks to fit the shape of the desired segmentation region. We comprehensively evaluated the proposed method using multiple datasets under varying conditions. The results show that the proposed method outperforms existing methods and provides a more detailed segmentation.

We present a reconfigurable deep learning pedestrian detection system for surveillance systems that detect people with shadows in different lighting and heavily occluded conditions. This work proposes a region-based CNN, combined with CMOS and thermal cameras to obtain human features even under poor lighting conditions. The main advantage of a reconfigurable system with respect to processor-based systems is its high performance and parallelism when processing large amount of data such as video frames. We discuss the details of hardware implementation in the proposed real-time pedestrian detection algorithm on a Zynq FPGA. Simulation results show that the proposed integrated approach of R-CNN architecture with cameras provides better performance in terms of accuracy, precision, and F1-score. The performance of Zynq FPGA was compared to other works, which showed that the proposed architecture is a good trade-off in terms of quality, accuracy, speed, and resource utilization.

We propose Quality Enhancement via a Side bitstream Network (QESN) technique for lossy image compression. The proposed QESN utilizes the network architecture of deep image compression to produce a bitstream for enhancing the quality of conventional compression. We also present a loss function that directly optimizes the Bjontegaard delta bit rate (BD-BR) by using a differentiable model of a rate-distortion curve. Experimental results show that QESN improves the rate by 16.7% in the BD-BR compared to Better Portable Graphics.

In CRYPTO 2019, Gohr first introduced the deep learning method to cryptanalysis for SPECK32/64. A differential-neural distinguisher was obtained using ResNet neural network. Zhang et al. used multiple parallel convolutional layers with different kernel sizes to capture information from multiple dimensions, thus improving the accuracy or obtaining a more round of distinguisher for SPECK32/64 and SIMON32/64. Inspired by Zhang's work, we apply the network structure to other ciphers. We not only improve the accuracy of the distinguisher, but also increase the number of rounds of the distinguisher, that is, distinguish more rounds of ciphertext and random number for DES, Chaskey and PRESENT.

This paper proposes a deep neural network named BayesianPUFNet that can achieve high prediction accuracy even with few challenge-response pairs (CRPs) available for training. Generally, modeling attacks are a vulnerability that could compromise the authenticity of physically unclonable functions (PUFs); thus, various machine learning methods including deep neural networks have been proposed to assess the vulnerability of PUFs. However, conventional modeling attacks have not considered the cost of CRP collection and analyzed attacks based on the assumption that sufficient CRPs were available for training; therefore, previous studies may have underestimated the vulnerability of PUFs. Herein, we show that the application of Bayesian deep neural networks that incorporate Bayesian statistics can provide accurate response prediction even in situations where sufficient CRPs are not available for learning. Numerical experiments show that the proposed model uses only half the CRP to achieve the same response prediction as that of the conventional methods. Our code is openly available on https://github.com/bayesian-puf-net/bayesian-puf-net.git.

The original single-shot multibox detector (SSD) algorithm has good detection accuracy and speed for regular object recognition. However, the SSD is not suitable for detecting small objects for two reasons: 1) the relationships among different feature layers with various scales are not considered, 2) the predicted results are solely determined by several independent feature layers. To enhance its detection capability for small objects, this study proposes an improved SSD-based algorithm called proportional channels' fusion SSD (PCF-SSD). Three enhancements are provided by this novel PCF-SSD algorithm. First, a fusion feature pyramid model is proposed by concatenating channels of certain key feature layers in a given proportion for object detection. Second, the default box sizes are adjusted properly for small object detection. Third, an improved loss function is suggested to train the above-proposed fusion model, which can further improve object detection performance. A series of experiments are conducted on the public database Pascal VOC to validate the PCF-SSD. On comparing with the original SSD algorithm, our algorithm improves the mean average precision and detection accuracy for small objects by 3.3% and 3.9%, respectively, with a detection speed of 40FPS. Furthermore, the proposed PCF-SSD can achieve a better balance of detection accuracy and efficiency than the original SSD algorithm, as demonstrated by a series of experimental results.

Tool condition monitoring is one of the core tasks of intelligent manufacturing in digital workshop. This paper presents an intelligent recognize method of tool condition based on deep learning. First, the industrial microphone is used to collect the acoustic signal during machining; then, a central fractal decomposition algorithm is proposed to extract sensitive information; finally, the multi-scale convolutional recurrent neural network is used for deep feature extraction and pattern recognition. The multi-process milling experiments proved that the proposed method is superior to the existing methods, and the recognition accuracy reached 88%.

Automatic continuous tracking of objects involved in a construction project is required for such tasks as productivity assessment, unsafe behavior recognition, and progress monitoring. Many computer-vision-based tracking approaches have been investigated and successfully tested on construction sites; however, their practical applications are hindered by the tracking accuracy limited by the dynamic, complex nature of construction sites (i.e. clutter with background, occlusion, varying scale and pose). To achieve better tracking performance, a novel deep-learning-based tracking approach called the Multi-Domain Convolutional Neural Networks (MD-CNN) is proposed and investigated. The proposed approach consists of two key stages: 1) multi-domain representation of learning; and 2) online visual tracking. To evaluate the effectiveness and feasibility of this approach, it is applied to a metro project in Wuhan China, and the results demonstrate good tracking performance in construction scenarios with complex background. The average distance error and F-measure for the MDNet are 7.64 pixels and 67, respectively. The results demonstrate that the proposed approach can be used by site managers to monitor and track workers for hazard prevention in construction sites.