Search

Deep learning technology has grown significantly in new application scenarios such as smart cities and driverless vehicles, but its deployment needs to consume a lot of resources. It is usually difficult to execute inference task solely on resource-constrained Intelligent Internet-of-Things (IoT) devices to meet strictly service delay requirements. CNN-based inference task is usually offloaded to the edge server or cloud. However, it may lead to unstable performance and privacy leaks. To address the above challenges, this article aims to design a low latency distributed inference framework, EdgeCI, which assigns inference tasks to locally idle, connected, and resource-constrained IoT device cluster networks. EdgeCI exploits two key optimization knobs, including: (1) Auction-based Workload Assignment Scheme (AWAS), which achieves the workload balance by assigning each workload partition to the more matching IoT device; (2) Fused-Layer parallelization strategy based on non-recursive Dynamic Programming (DPFL), which is aimed at further minimizing the inference time. We have implemented EdgeCI based on PyTorch and evaluated its performance with VGG-16 and ResNet-34 image recognition models. The experimental results prove that our proposed AWAS and DPFL outperform the typical state-of-the-art solutions. When they are well combined, EdgeCI can improve inference speed by 34.72% to 43.52%. EdgeCI outperforms the state-of-the art approaches on our edge cluster. [ABSTRACT FROM AUTHOR]

3D perception of depth and ego-motion is of vital importance in intelligent agent and Human Computer Interaction (HCI) tasks, such as robotics and autonomous driving. There are different kinds of sensors that can directly obtain 3D depth information. However, the commonly used Lidar sensor is expensive, and the effective range of RGB-D cameras is limited. In the field of computer vision, researchers have done a lot of work on 3D perception. While traditional geometric algorithms require a lot of manual features for depth estimation, Deep Learning methods have achieved great success in this field. In this work, we proposed a novel self-supervised method based on Vision Transformer (ViT) with Convolutional Neural Network (CNN) architecture, which is referred to as ViT-Depth. The image reconstruction losses computed by the estimated depth and motion between adjacent frames are treated as supervision signal to establish a self-supervised learning pipeline. This is an effective solution for tasks that need accurate and low-cost 3D perception, such as autonomous driving, robotic navigation, 3D reconstruction, and so on. Our method could leverage both the ability of CNN and Transformer to extract deep features and capture global contextual information. In addition, we propose a cross-frame loss that could constrain photometric error and scale consistency among multi-frames, which lead the training process to be more stable and improve the performance. Extensive experimental results on autonomous driving dataset demonstrate the proposed approach is competitive with the state-of-the-art depth and motion estimation methods. [ABSTRACT FROM AUTHOR]

Delta compression, as a complementary technique for data deduplication, has gained widespread attention in network storage systems. It can eliminate redundant data between non‐duplicate but similar chunks that cannot be identified by data deduplication. The network transmission overhead between servers and clients can be greatly reduced by using data deduplication and delta compression techniques. Resemblance detection is a technique that identifies similar chunks for post‐deduplication delta compression in network storage systems. The existing resemblance detection approaches fail to detect similar chunks with arbitrary similarity by setting a similarity threshold, which can be suboptimal. In this paper, the authors propose Chunk2vec, a resemblance detection scheme for delta compression that utilizes deep learning techniques and Approximate Nearest Neighbour Search technique to detect similar chunks with any given similarity range. Chunk2vec uses a deep neural network, Sentence‐BERT, to extract an approximate feature vector for each chunk while preserving its similarity with other chunks. The experimental results on five real‐world datasets indicate that Chunk2vec improves the accuracy of resemblance detection for delta compression and achieves higher compression ratio than the state‐of‐the‐art resemblance detection technique. [ABSTRACT FROM AUTHOR]

With the development of deep learning, the demand for similarity matching between texts in text classification is becoming increasingly high. How to match texts quickly under the premise of keeping private information secure has become a research hotspot. However, most existing protocols currently have full set limitations, and the applicability of these methods is limited when the data size is large and scattered. Therefore, this paper applies the secure vector calculation method for text similarity matching in the case of data without any complete set constraints, and it designs a secure computation protocol of text similarity (SCTS) based on the semi-honest model. At the same time, elliptic-curve cryptography technology is used to greatly improve the execution efficiency of the protocol. In addition, we also analyzed the possibility of the malicious behavior of participants in the semi-honest-model protocol, and further designed an SCTS protocol suitable for the malicious model using the cut-and-choose and zero-knowledge-proof methods. By proposing a security mechanism, this protocol aims to provide a reliable and secure computing solution that can effectively prevent malicious attacks and interference. Finally, through the analysis of the efficiencies of the existing protocols, the efficiencies of the protocols under the malicious model are further verified, and the practical value for text classification in deep learning is demonstrated. [ABSTRACT FROM AUTHOR]

With natural language processing as an important research direction in deep learning, the problems of text similarity calculation, natural language inference, question and answer systems, and information retrieval can be regarded as text matching applications for different data and scenarios. Secure matching computation of text string patterns can solve the privacy protection problem in the fields of biological sequence analysis, keyword search, and database query. In this paper, we propose an Intelligent Semi-Honest System (ISHS) for secret matching against malicious adversaries. Firstly, a secure computation protocol based on the semi-honest model is designed for the secret matching of text strings, which adopts a new digital encoding method and an ECC encryption algorithm and can provide a solution for honest participants. The text string matching protocol under the malicious model which uses the cut-and-choose method and zero-knowledge proof is designed for resisting malicious behaviors that may be committed by malicious participants in the semi-honest protocol. The correctness and security of the protocol are analyzed, which is more efficient and has practical value compared with the existing algorithms. The secure text matching has important engineering applications. [ABSTRACT FROM AUTHOR]

At present, the amount of network equipment, servers, and network traffic is increasing exponentially, and the way in which operators allocate and efficiently utilize network resources has attracted considerable attention from traffic forecasting researchers. However, with the advent of the 5G era, network traffic has also shown explosive growth, and network complexity has increased dramatically. Accurately predicting network traffic has become a pressing issue that must be addressed. In this paper, a multilayer perceptron ensemble learning method based on convolutional neural networks (CNN) and gated recurrent units (GRU) spatiotemporal feature extraction (MECG) is proposed for network traffic prediction. First, we extract spatial and temporal features of the data by convolutional neural networks (CNN) and recurrent neural networks (RNN). Then, the extracted temporal features and spatial features are fused into new spatiotemporal features through integrated learning of a multilayer perceptron, and a spatiotemporal prediction model is built in the sequence-to-sequence framework. At the same time, the teacher forcing mechanism and attention mechanism are added to improve the accuracy and convergence speed of the model. Finally, the proposed method is compared with other deep learning models for experiments. The experimental results show that the proposed method not only has apparent advantages in accuracy but also shows some superiority in time training cost. [ABSTRACT FROM AUTHOR]

Featured Application: Facial recognition. Automated facial gender and age classification has remained a challenge because of the high inter-subject and intra-subject variations. We addressed this challenging problem by studying multi-instance- and multi-scale-enhanced multi-task random forest architecture. Different from the conventional single facial attribute recognition method, we designed effective multi-task architecture to learn gender and age simultaneously and used the dependency between gender and age to improve its recognition accuracy. In the study, we found that face gender has a great influence on face age grouping; thus, we proposed a random forest face age grouping method based on face gender conditions. Specifically, we first extracted robust multi-instance and multi-scale features to reduce the influence of various intra-subject distortion types, such as low image resolution, illumination and occlusion, etc. Furthermore, we used a random forest classifier to recognize facial gender. Finally, a gender conditional random forest was proposed for age grouping to address inter-subject variations. Experiments were conducted by using two popular MORPH-II and Adience datasets. The experimental results showed that the gender and age recognition rates in our method can reach 99.6% and 96.14% in the MORPH-II database and 93.48% and 63.72% in the Adience database, reaching the state-of-the-art level. [ABSTRACT FROM AUTHOR]

Aortic dissection (AD) is one of the most serious diseases with high mortality, and its diagnosis mainly depends on computed tomography (CT) results. Most existing automatic diagnosis methods of AD are only suitable for AD recognition, which usually require preselection of CT images and cannot be further classified to different types. In this work, we constructed a dataset of 105 cases with a total of 49021 slices, including 31043 slices expertlevel annotation and proposed a two-stage AD diagnosis structure based on sequence information and deep learning. The proposed region of interest (RoI) extraction algorithm based on sequence information (RESI) can realize high-precision for RoI identification in the first stage. Then DenseNet-121 is applied for further diagnosis. Specially, the proposed method can judge the type of AD without preselection of CT images. The experimental results show that the accuracy of Stanford typing classification of AD is 89.19%, and the accuracy at the slice-level reaches 97.41%, which outperform the state-ofart methods. It can provide important decision-making information for the determination of further surgical treatment plan for patients. [ABSTRACT FROM AUTHOR]

Automated industrial quality detection (QD) boosts quality-detection efficiency and reduces costs. However, current quality-detection algorithms have drawbacks such as low efficiency, easily missed detections, and false detections. We propose QD-YOLO, an attention-based method to enhance quality-detection efficiency on computer mainboards. Firstly, we propose a composite attention module for the network's backbone to highlight appropriate feature channels and improve the feature fusion structure, allowing the network to concentrate on the crucial information in the feature map. Secondly, we employ the Meta-ACON activation function to dynamically learn whether the activation function is linear or non-linear for various input data and adapt it to varied input scenarios with varying linearity. Additionally, we adopt Ghost convolution instead of ordinary convolution, using linear operations as possible to reduce the number of parameters and speed up detection. Experimental results show that our method can achieve improved real-time performance and accuracy on the self-created mainboard quality defect dataset, with a mean average precision (mAP) of 98.85% and a detection speed of 31.25 Frames Per Second (FPS). Compared with the original YOLOv5s model, the improved method improves mAP@0.5 by 2.09% and detection speed by 2.67 FPS. [ABSTRACT FROM AUTHOR]

At present, solid-state fermentation (SSF) is mainly controlled by artificial experience, and the product quality and yield are not stable. Therefore, predicting the quality and yield of SSF is of great significance for improving the utility of SSF. In this article, we propose a deep learning utility prediction (DLUP) scheme for the SSF in the Industrial Internet of Things, including parameters collection and utility prediction of the SSF process. Furthermore, we propose a novel edge-rewritable Petri net to model the parameters collection and utility prediction of the SSF process and further verify their soundness. More importantly, DLUP combines the generating ability of least squares generative adversarial network with the predicting ability of fully connected neural network to realize the utility prediction (usually use the alcohol concentration) of SSF. Experiments show that the proposed method predicts the alcohol concentration more accurately than the other joint prediction methods. In addition, the method in our article provides evidences for setting the ratio of raw materials and proper temperature through numerical analysis. [ABSTRACT FROM AUTHOR]

Internet of Things (IoT) realizes the interconnection of heterogeneous devices by the technology of wireless and mobile communication. The data of target regions are collected by widely distributed sensing devices and transmitted to the processing center for aggregation and analysis as the basis of IoT. The quality of IoT services usually depends on the accuracy and integrity of data. However, due to the adverse environment or device defects, the collected data will be anomalous. Therefore, the effective method of anomaly detection is the crucial issue for guaranteeing service quality. Deep learning is one of the most concerned technology in recent years which realizes automatic feature extraction from raw data. In this article, the integrated model of the convolutional neural network (CNN) and recurrent autoencoder is proposed for anomaly detection. Simple combination of CNN and autoencoder cannot improve classification performance, especially, for time series. Therefore, we utilize the two-stage sliding window in data preprocessing to learn better representations. Based on the characteristics of the Yahoo Webscope S5 dataset, raw time series with anomalous points are extended to fixed-length sequences with normal or anomaly label via the first-stage sliding window. Then, each sequence is transformed into continuous time-dependent subsequences by another smaller sliding window. The preprocessing of the two-stage sliding window can be considered as low-level temporal feature extraction, and we empirically prove that the preprocessing of the two-stage sliding window will be useful for high-level feature extraction in the integrated model. After data preprocessing, spatial and temporal features are extracted in CNN and recurrent autoencoder for the classification in fully connected networks. Empiric results show that the proposed model has better performances on multiple classification metrics and achieves preferable effect on anomaly detection. [ABSTRACT FROM AUTHOR]

Medical image segmentation is an important application field of computer vision in medical image processing. Due to the close location and high similarity of different organs inmedical images, the current segmentation algorithms have problems with mis-segmentation and poor edge segmentation. To address these challenges, we propose a medical image segmentation network (AF-Net) based on attention mechanism and feature fusion, which can effectively capture global information while focusing the network on the object area. In this approach, we add dual attention blocks (DA-block) to the backbone network, which comprises parallel channels and spatial attention branches, to adaptively calibrate and weigh features. Secondly, the multi-scale feature fusion block (MFF-block) is proposed to obtain feature maps of different receptive domains and get multi-scale informationwith less computational consumption. Finally, to restore the locations and shapes of organs, we adopt the global feature fusion blocks (GFF-block) to fuse high-level and low-level information, which can obtain accurate pixel positioning. We evaluate our method on multiple datasets(the aorta and lungs dataset), and the experimental results achieve 94.0% in mIoU and 96.3% in DICE, showing that our approach performs better than U-Net and other state-of-art methods. [ABSTRACT FROM AUTHOR]

In recent years, with the development of machine learning and deep learning, it is possible to identify and even control crop diseases by using electronic devices instead of manual observation. In this paper, an image recognition method of citrus diseases based on deep learning is proposed. We built a citrus image dataset including six common citrus diseases. The deep learning network is used to train and learn these images, which can effectively identify and classify crop diseases. In the experiment, we use MobileNetV2 model as the primary network and compare it with other network models in the aspect of speed, model size, accuracy. Results show that our method reduces the prediction time consumption and model size while keeping a good classification accuracy. Finally, we discuss the significance of using MobileNetV2 to identify and classify agricultural diseases in mobile terminal, and put forward relevant suggestions. [ABSTRACT FROM AUTHOR]

Simple Summary: Visual identification of cattle in a realistic farming environment is helpful for real-time cattle monitoring. Based on continuous cattle detection, identification, and behavior recognition, it is possible to utilize cameras on farms within company or government networks to provide the services of production supervision, early disease detection, and animal science research for precision livestock farming. However, cattle identification in the wild is still a difficult problem due to the high similarities of different identities and the variances of the same identity as posture or perspective changes. Our proposed method based on deep convolutional neural networks and deep metric learning provides a promising approach for cattle identification and paves the way toward continuous monitoring of cattle in a nearly natural state. Visual identification of cattle in the wild provides an essential way for real-time cattle monitoring applicable to precision livestock farming. Chinese Simmental exhibit a yellow or brown coat with individually characteristic white stripes or spots, which makes a biometric identifier for identification possible. This work employed the observable biometric characteristics to perform cattle identification with an image from any viewpoint. We propose multi-center agent loss to jointly supervise the learning of DCNNs by SoftMax with multiple centers and the agent triplet. We reformulated SoftMax with multiple centers to reduce intra-class variance by offering more centers for feature clustering. Then, we utilized the agent triplet, which consisted of the features and the agents, to enforce separation among different classes. As there are no datasets for the identification of cattle with multi-view images, we created CNSID100, consisting of 11,635 images from 100 Chinese Simmental identities. Our proposed loss was comprehensively compared with several well-known losses on CNSID100 and OpenCows2020 and analyzed in an engineering application in the farming environment. It was encouraging to find that our approach outperformed the state-of-the-art models on the datasets above. The engineering application demonstrated that our pipeline with detection and recognition is promising for continuous cattle identification in real livestock farming scenarios. [ABSTRACT FROM AUTHOR]

A knowledge graph (KG), also known as a knowledge base, is a particular kind of network structure in which the node indicates entity and the edge represent relation. However, with the explosion of network volume, the problem of data sparsity that causes large-scale KG systems to calculate and manage difficultly has become more significant. For alleviating the issue, knowledge graph embedding is proposed to embed entities and relations in a KG to a low-, dense and continuous feature space, and endow the yield model with abilities of knowledge inference and fusion. In recent years, many researchers have poured much attention in this approach, and we will systematically introduce the existing state-of-the-art approaches and a variety of applications that benefit from these methods in this paper. In addition, we discuss future prospects for the development of techniques and application trends. Specifically, we first introduce the embedding models that only leverage the information of observed triplets in the KG. We illustrate the overall framework and specific idea and compare the advantages and disadvantages of such approaches. Next, we introduce the advanced models that utilize additional semantic information to improve the performance of the original methods. We divide the additional information into two categories, including textual descriptions and relation paths. The extension approaches in each category are described, following the same classification criteria as those defined for the triplet fact-based models. We then describe two experiments for comparing the performance of listed methods and mention some broader domain tasks such as question answering, recommender systems, and so forth. Finally, we collect several hurdles that need to be overcome and provide a few future research directions for knowledge graph embedding. [ABSTRACT FROM AUTHOR]