Search

X‐ray security checks aim to detect contraband in luggage; however, the detection accuracy is hindered by the overlapping and significant size differences of objects in X‐ray images. To address these challenges, the authors introduce a novel network model named Multi‐Scale Feature Attention (MSFA)‐DEtection TRansformer (DETR). Firstly, the pyramid feature extraction structure is embedded into the self‐attention module, referred to as the MSFA. Leveraging the MSFA module, MSFA‐DETR extracts multi‐scale feature information and amalgamates them into high‐level semantic features. Subsequently, these features are synergised through attention mechanisms to capture correlations between global information and multi‐scale features. MSFA significantly bolsters the model's robustness across different sizes, thereby enhancing detection accuracy. Simultaneously, A new initialisation method for object queries is proposed. The authors' foreground sequence extraction (FSE) module extracts key feature sequences from feature maps, serving as prior knowledge for object queries. FSE expedites the convergence of the DETR model and elevates detection accuracy. Extensive experimentation validates that this proposed model surpasses state‐of‐the‐art methods on the CLCXray and PIDray datasets. [ABSTRACT FROM AUTHOR]

Due to the potential for airborne radar to capture incomplete observational information regarding unmanned aerial vehicle (UAV) trajectories, this study introduces a novel approach called Node-former, which integrates neural ordinary differential equations (NODEs) and the Informer framework. The proposed method exhibits high accuracy in trajectory prediction, even in scenarios with prolonged data interruptions. Initially, data outside the acceptable error range are discarded to mitigate the impact of interruptions on prediction accuracy. Subsequently, to address the irregular sampling caused by data elimination, NODEs are utilized to transform computational interpolation into an initial value problem (IPV), thus preserving informative features. Furthermore, this study enhances the Informer's encoder through the utilization of time-series prior knowledge and introduces an ODE solver as the decoder to mitigate fluctuations in the original decoder's output. This approach not only accelerates feature extraction for long sequence data, but also ensures smooth and robust output values. Experimental results demonstrate the superior performance of Node-former in trajectory prediction with interrupted data compared to traditional algorithms. [ABSTRACT FROM AUTHOR]

Accurate preoperative microvascular invasion (MVI) prediction using PET/CT imaging is important for guiding surgical plans and treatment strategies for hepatocellular carcinoma (HCC) patients. Deep learning (DL)-based approaches have shown effectiveness in medical imaging-based diagnostics. However, most DL-based methods for automatic MVI diagnosis depend on detailed annotations of HCC regions by clinicians, then perform feature extraction and directly output MVI risk scores, which can increase application complexity and lack transparency in decision-making. Therefore, we propose the sub-region aware retrieval-based network (SRANet), utilizing the prior knowledge of modality complementarity and localization of HCC in PET/CT images for tumor area location and feature extraction. Specifically, in the feature extraction stage, to better highlight lesion areas, we designed the prior knowledge guided fusion (PKGF) block. PKGF integrates knowledge from PET and CT through attention mechanism-based intermodal interaction and implicitly weights features to calibrate different modalities. Furthermore, we developed a sub-region aware module (SRAM) that employs class activation schemes and peak selection to identify and highlight sparse, salient sub-regions in tumor areas, relying solely on coarse-grained supervision. Additionally, we design a new loopback Siamese training strategy, guiding the model to focus on lesion areas through the semantic invariance of lesion features. Finally, we employ a retrieval strategy to present similar slices, providing clinicians with transparent references and enhancing the understandability of MVI predictions beyond simple risk scores. Experimental results show SRANet outperforms leading methods, achieving a slice-level AUC of 0.855 and an individual-level AUC of 0.796, highlighting its potential for clinical application. [ABSTRACT FROM AUTHOR]

This paper presents an algorithm and code for training a variational autoencoder to perform semantic transformation of USDA SR legacy food data into latent space. The term transformation is used rather than dimensionality reduction to emphasize that the main aim of the network is to learn to features from it, apart from compressing. The model is expected to learn features (glycaemic index, superfood), and filter out shallow attributes for lossless compression, which would ease the training of any other neural network it. The dimension has been reduced by 50% as a byproduct. The architecture designed to be equipping with prior knowledge [13]. The scope of this study is to realize this goal with minimum resources and maximum accuracy. Dataset was limited and diverse, which was challenging to train. Traditional dimensionality reduction techniques do not promise extraction of useful feature that condemns the scalability of the project from further application. [ABSTRACT FROM AUTHOR]

Multi-label Zero-shot Learning (ZSL) is more reasonable and realistic than standard single-label ZSL because several objects can co-exist in a natural image in real scenarios. Intra-class feature entanglement is a significant factor influencing the alignment of visual and semantic features, resulting in the model's inability to recognize unseen samples comprehensively and completely. We observe that existing multi-label ZSL methods place a greater emphasis on attention-based refinement and decoupling of visual features, while ignoring the relationship between label semantics. Relying on label correlations to solve multi-label ZSL tasks has not been deeply studied. In this paper, we make full use of the co-occurrence relationship between category labels and build a directed weighted semantic graph based on statistics and prior knowledge, in which node features represent category semantics and weighted edges represent conditional probabilities of label co-occurrence. To guide the targeted extraction of visual features, node features and edge set weights are simultaneously updated and refined, and embedded into the visual feature extraction network from a global and local perspective. The proposed method's effectiveness was demonstrated by simulation results on two challenging multi-label ZSL benchmarks: NUS-WIDE and Open Images. In comparison to state-of-the-art models, our model achieves an absolute gain of 2.4% mAP on NUS-WIDE and 2.1% mAP on Open Images respectively. • Semantic node relationships are represented by directed weighted graphs. • A simultaneous update strategy is proposed to refine the semantic graph in an all-round way. • Semantic and visual features are embedded and interacted with local and global response ways at different stages. • Extensive experimental results demonstrate the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

Seismic surveys are often affected by environmental obstacles or restrictions that prevent regular sampling in seismic acquisition. To address missing data, various methods, including deep learning techniques, have been developed to extract features from complex information, albeit with the limitation of requiring external seismic databases. While previous works have primarily focused on trace reconstruction, missing shot-gathers directly impact the seismic processing flow and represent a major challenge in seismic data regularization. In this paper, we propose DIPsgr, a seismic shot-gather reconstruction method that uses only the incomplete seismic acquisition measurements to estimate their missing information employing unsupervised deep learning. Numerical experiments on three databases demonstrate that DIPsgr recovers the complete set of traces in each shot-gather, with preserved information and seismic events. [ABSTRACT FROM AUTHOR]

Efficient object detection and tracking from remote sensing video data acquired by unmanned aerial vehicles (UAVs) has significant implications in various domains, such as scene understanding, traffic surveillance, and military operations. Although the modern transformer-based trackers have demonstrated superior tracking accuracy, they often require extensive training time to achieve convergence, and the information from templates is not fully utilized in and integrated into tracking. To accelerate convergence and further improve tracking accuracy, we propose an end-to-end tracker named ParallelTracker that extracts prior knowledge from templates for better convergence and enriches template features in a parallel manner. Our core design incorporates spatial prior knowledge into the tracking process through three modules: a prior knowledge extractor module (PEM), a template features parallel enhancing module (TPM), and a template prior knowledge merge module (TPKM). These modules enable rich and discriminative feature extraction as well as integration of target information. We employ multiple PEM, TPM and TPKM modules along with a localization head to enhance accuracy and convergence speed in object tracking. To enable efficient online tracking, we also design an efficient parallel scoring prediction head (PSH) for selecting high-quality online templates. Our ParallelTracker achieves state-of-the-art performance on the UAV tracking benchmark UAV123, with an AUC score of 69.29%, surpassing the latest OSTrack and STARK methods. Ablation studies further demonstrate the positive impact of our designed modules on both convergence and accuracy. [ABSTRACT FROM AUTHOR]

There are problems associated with facial expression recognition (FER), such as facial occlusion and head pose variations. These two problems lead to incomplete facial information in images, making feature extraction extremely difficult. Most current methods use prior knowledge or fixed-size patches to perform local cropping, thereby enhancing the ability to acquire fine-grained features. However, the former requires extra data processing work and is prone to errors; the latter destroys the integrity of local features. In this paper, we propose a local Sliding Window Attention Network (SWA-Net) for FER. Specifically, we propose a sliding window strategy for feature-level cropping, which preserves the integrity of local features and does not require complex preprocessing. Moreover, the local feature enhancement module mines fine-grained features with intraclass semantics through a multiscale depth network. The adaptive local feature selection module is introduced to prompt the model to find more essential local features. Extensive experiments demonstrate that our SWA-Net model achieves a comparable performance to that of state-of-the-art methods with scores of 90.03% on RAF-DB, 89.22% on FERPlus, 63.97% on AffectNet. [ABSTRACT FROM AUTHOR]

Underwater object detection is one of the important technologies for improving the efficiency of underwater inspection, but the existing methods still suffer from the problems of missed detection and insufficient target localization capability of targets. To address these problems, an improved Transformer and multi-scale attentional supervised feature fusion-based underwater object detection method is proposed. In our method, the underwater objects are preprocessed by prior knowledge first. Then, a new coordinate decomposition window-based (CDW) Transformer block is proposed to extract spatial location information more accurately, and scaling factors are introduced to reduce the intermediate computation. Finally, an attentional supervised fusion (ASF) method is proposed to strengthen the link between feature extraction and feature fusion, and further improve the detected performance by using compound attention weights. The cascade detection head is improved, where the information flow is reversed to enhance the prediction of coordinates. The average accuracy of the proposed method on the URPC and DUO datasets is 3.7% and 3.8% higher than that of the baseline network through the cross-test, and outperforms the state-of-the-art methods. This study can provide a reference for engineering applications such as automated marine operations and biodetected fishing techniques. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Computer Engineering & Applications is the property of Beijing Journal of Computer Engineering & Applications Journal Co Ltd. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Data-driven intelligent diagnosis models expect to mine the health information of machines from massive monitoring data. However, the size of faulty monitoring data collected in engineering scenarios is limited, which leads to few-shot fault diagnosis as a valuable research point. Fortunately, it is possible to reduce the required amount of training data by integrating prior diagnosis knowledge into diagnosis models. Inspired by this, we present a prior knowledge-augmented self-supervised feature learning framework for few-shot fault diagnosis. In the framework, 24 signal feature indicators are built to form prior features set based on existing diagnosis knowledge. Besides, a convolutional autoencoder is used to mine the general features, which are considered to potentially contain fault information that prior features do not possess. We design a self-supervised learning scheme for training the diagnosis model, which enables the model to learn both prior and general features served as proxy labels. As a result, the model is expected to mine richer features from limited monitoring data. The effectiveness of the proposed framework is verified using two mechanical fault simulation experiments. From the angle of prior diagnosis knowledge, the proposed framework provides a new perspective to the problem of few-shot intelligent diagnosis of machines. [ABSTRACT FROM AUTHOR]

Extracting intuitive operating state features from vibration signals without prior knowledge is a prospective requirement for health monitoring and fault diagnosis in bearings. In this paper, a visualized stacked denoising auto-encoder (VSDAE) model is proposed for the unsupervised extraction and quantitative evaluation of bearings' state features. First, the stacked denoising auto-encoder (SDAE) was used to reconstruct vibration signals. The intermediate vector of the SDAE, which is a high-information-density representation of vibration signals, was regarded as the pending state feature. Then, the dimension of the intermediate vector was reduced by the t-distributed stochastic neighbor embedding (t-SNE) method to the two-dimensional visualization space. Finally, the silhouette coefficient of feature distribution was calculated to quantitatively evaluate the extracted features. The proposed model was evaluated using experimental bearing signals simulating various operating states. The results proved that the features, extracted and evaluated by the VSDAE, allowed the recognition of the operating states of the examined bearings. [ABSTRACT FROM AUTHOR]

• A novel perspective on anomaly detection along the coefficient domain is explored. • The discriminative coefficient distribution pattern between background and anomalies is revealed. • The negative impact of potential anomalies on CRD is effectively eliminated. Various anomaly detection (AD) methods focus on the background feature extraction and suppression from hyperspectral images (HSIs). However, this process is susceptible to anomaly signals. As a typical example, the collaborative representation (CR) model incorporates contextual information into a linear approximation of the central pixel, but suffers severely from anomaly contamination from the local neighborhood. How to remove potential anomalies in the complete lack of prior knowledge is a crucial problem, yet there is currently no reliable solution due to the limited background-anomaly separation in the original HSI space. In this paper, inspired by the unique coefficient distribution in the coefficient domain, a discriminative coefficient analysis (DCA)-based collaborative representation method is proposed for hyperspectral AD. By projecting the HSI into a local coefficient domain relative to the central pixel, an interesting observation occurs that the coefficients of dominant background atoms exhibit a smooth and stable distribution pattern, while those of few anomalous atoms deviate significantly from the mainstream. Accordingly, by designing a pair of background coefficient boundaries, neighboring atoms with coefficient lying outside are considered suspicious anomalies and thus removed. The remaining atoms, corresponding to one homogeneous or few categories of local background materials, finally contribute to accurate anomaly judgment on the central pixel. Different from existing studies, the main contribution lies in the novel perspective of discriminative coefficient domain with enhanced background-anomaly separation. Experiments on one synthetic and five real-world HSI data sets demonstrate the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

In industrial systems, textual failure records note the failure mechanisms, the parts involved, and the failure symptoms; these records guide fault analysis and repair. However, case retrieval and feature extraction require extensive prior knowledge and diagnostic expertise; this method is time-consuming and labor-intensive. In this article, we present a novel two-stage framework that automatically extracts and generates features from very large textual records. We use an improved weighted latent Dirichlet allocation model and the Word2vec method to extract topic category and semantic features from fault texts; this approach accelerates training convergence. Next, we build a topic-context attention model in which word-embedding semantic features interact with topic features. Finally, we use classification and similarity calculation models to diagnose faults and retrieve similar cases; this approach ensures feature generation. Our method is very granular in terms of case representation which significantly improves diagnosis. The method robustly identifies similar cases by interrogating vehicle maintenance records. [ABSTRACT FROM AUTHOR]

With the widespread use of deep learning in single object tracking task, mainstream tracking algorithms treat tracking as a combined classification and regression problem. Classification aims at locating an arbitrary target, and regression aims at estimating the corresponding bounding box. In this paper, we focus on regression and propose a novel box estimation network, which consists of a transformer encoder target pyramid guide (TPG) and transformer decoder target pyramid spread (TPS). Specifically, the transformer encoder TPG is designed to generate fine-grained prior knowledge with explicit representation for template targets. In contrast to the raw transformer encoder, we capture the visual dependence through local-global self-attention and deem the multi-scale target regions as the “local” region. Using this fine-grained prior knowledge, we design the transformer decoder TPS to spread it to the subsequent search regions with high affinity to accurately estimate the bounding boxes. Considering that self-attention fails to model information interaction across channels between the template target and search regions, we develop a channel-wise cross-attention block within the TPS as compensation. Extensive experiments on the OTB100, UAV123, NFS, VOT2020, VOT2021, LaSOT, LaSOT_ext, TrackingNet and GOT-10k benchmarks show that the proposed box estimation network outperforms most existing box estimation methods. Furthermore, our trackers based on this estimation network exhibit a competitive performance against state-of-the-art trackers. [ABSTRACT FROM AUTHOR]

Detecting the small targets from a heterogeneous background in an infrared image is a challenging problem, which has received extensive attention. In this article, we propose a method in terms of tensor ring (TR) decomposition and nonlinear multiscale morphological top-hat transformation for infrared small target detection (ISTD). First, a tensor model with prior knowledge is constructed for extracting the structural features of multiple infrared images. Then, the problem of small target detection is converted into a problem of minimizing the tensor rank with TR. Based on the TR decomposition model, we introduce the top-hat regularization into our model with multiple structural elements of different size to perform morphological operations. The corresponding morphological model exploits a more accurate ring top-hat regularization expression through adaptive nonlinear combination for the ISTD problem. Finally, the optimization of the model is realized by the closed solution given by the alternating direction method of multipliers algorithm. In order to verify the superior performance of our method, our method is compared with a number of advanced detection models. By analyzing the results of comparison experiments, the detection accuracy and precision of our model in the detection of small infrared targets have been improved. Even in complex background conditions, our model also maintain a good robustness. [ABSTRACT FROM AUTHOR]

Text-to-image synthesis is a challenging task that generates realistic images from a textual description. The description contains limited information compared with the corresponding image and is ambiguous and abstract, which will complicate the generation and lead to low-quality images. To address this problem, we propose a novel generation text-to-image synthesis method, called RiFeGAN2, to enrich the given description. To improve the enrichment quality while accelerating the enrichment process, RiFeGAN2 exploits a domain-specific constrained model to limit the search scope and then uses an attention-based caption matching model to refine the compatible candidate captions based on constrained prior knowledge. To improve the semantic consistency between the given description and the synthesized results, RiFeGAN2 employs improved SAEMs, SAEM2s, to compact better features of the retrieved captions and effectively emphasize the descriptions via incorporating centre-attention layers. Finally, multi-caption attentional GANs are exploited to synthesize images from those features. Experiments performed on widely-used datasets show that the models can generate vivid images from enriched captions and effectually improve the semantic consistency. [ABSTRACT FROM AUTHOR]

Due to prior knowledge deficiency, large spectral variability, and high dimension of hyperspectral image (HSI), HSI clustering is extremally a fundamental but challenging task. Deep clustering methods have achieved remarkable success and have attracted increasing attention in unsupervised HSI classification (HSIC). However, the poor robustness, adaptability, and feature presentation limit their practical applications to complex large-scale HSI datasets. Thus, this article introduces a novel self-supervised locality preserving low-pass graph convolutional embedding method (L2GCC) for large-scale hyperspectral image clustering. Specifically, a spectral–spatial transformation HSI preprocessing mechanism is introduced to learn superpixel-level spectral–spatial features from HSI and reduce the number of graph nodes for subsequent network processing. In addition, locality preserving low-pass graph convolutional embedding autoencoder is proposed, in which the low-pass graph convolution and layerwise graph attention are designed to extract the smoother features and preserve layerwise locality features, respectively. Finally, we develop a self-training strategy, in which a self-training clustering objective employs soft labels to supervise the clustering process and obtain appropriate hidden representations for node clustering. L2GCC is an end-to-end training network, which is jointly optimized by graph reconstruction loss and self-training clustering loss. On Indian Pines, Salinas, and University of Houston 2013 datasets, the clustering accuracy overall accuracies (OAs) of the proposed L2GCC are 73.51%, 83.15%, and 64.12%, respectively. [ABSTRACT FROM AUTHOR]

Recent image classification efforts have achieved certain success by incorporating prior information such as labels and logical rules to learn discriminative features. However, these methods overlook the variability of features, resulting in feature inconsistency and fluctuations in model parameter updates, which further contribute to decreased image classification accuracy and model instability. To address this issue, this paper proposes a novel method combining structural prior-driven feature extraction with gradient-momentum (SPGM), from the perspectives of consistent feature learning and precise parameter updates, to enhance the accuracy and stability of image classification. Specifically, SPGM leverages a structural prior-driven feature extraction (SPFE) approach to calculate gradients of multi-level features and original images to construct structural information, which is then transformed into prior knowledge to drive the network to learn features consistent with the original images. Additionally, an optimization strategy integrating gradients and momentum (GMO) is introduced, dynamically adjusting the direction and step size of parameter updates based on the angle and norm of the sum of gradients and momentum, enabling precise model parameter updates. Extensive experiments on CIFAR10 and CIFAR100 datasets demonstrate that the SPGM method significantly reduces the top-1 error rate in image classification, enhances the classification performance, and outperforms state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Person re-identification (Re-ID) aims to retrieve the same person in the gallery. Transformers have been introduced to the Re-ID task due to their excellent ability to model long-range dependency. However, due to the properties of the global attention mechanism, they are less effective in capturing the discriminative local semantics of pedestrians compared to convolutional operations. To address this issue, we present a Multi-granularity Cross Transformer Network (MCTN) that progressively learns salient features of different local structures in a global context. Specifically, we first utilize a Multi-granularity Convolutional Layer (MCL) to investigate salient pedestrian features at various granularities. On this basis, we propose a Pyramidal Cross Transformer learning layer (PCT), which contains a pyramidal division of pedestrian image feature maps, differentiated feature extraction of different parts of pedestrians, and cross attention to exploring the local–global relationship of the feature map. It allows effective mining of local information in the global structure from a coarse-to-fine perspective. Furthermore, to enhance the interaction between low-level detailed features and high-level semantic features, a Hierarchical Aggregation Strategy (HAS) is introduced to fuse features learned by cross attention learning at different stages. Pedestrian features learned in shallow layers will serve as global priors for semantics learning in deep layers. We evaluate our method on four large-scale Re-ID datasets, and the experimental results reveal that the proposed method outperforms the state-of-the-art methods. • We propose a method that progressively learns discriminative semantic information. • The Pyramidal Cross Transformer Learning Layer obtains rich and diverse clues. • The Hierarchical Aggregation Strategy integrates low- and high-level information. • Extensive experimental results indicate the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

Part-level representations are important for robust person re-identification (ReID), but in practice feature quality suffers due to the body part misalignment problem. In this paper, we present a robust, compact, and easy-to-use method called the Multi-task Part-aware Network (MPN), which is designed to extract semantically aligned part-level features from pedestrian images. MPN solves the body part misalignment problem via multi-task learning (MTL) in the training stage. More specifically, it builds one main task (MT) and one auxiliary task (AT) for each body part on the top of the same backbone model. The ATs are equipped with a coarse prior of the body part locations for training images. ATs then transfer the concept of the body parts to the MTs via optimizing the MT parameters to identify part-relevant channels from the backbone model. Concept transfer is accomplished by means of two novel alignment strategies: namely, parameter space alignment via hard parameter sharing and feature space alignment in a class-wise manner. With the aid of the learned high-quality parameters, MTs can independently extract semantically aligned part-level features from relevant channels in the testing stage. MPN has three key advantages: 1) it does not need to conduct body part detection in the inference stage; 2) its model is very compact and efficient for both training and testing; 3) in the training stage, it requires only coarse priors of body part locations, which are easy to obtain. Systematic experiments on four large-scale ReID databases demonstrate that MPN consistently outperforms state-of-the-art approaches by significant margins. [ABSTRACT FROM AUTHOR]

• We propose a novel unmixing algorithm named spatial feature extraction non-negative tensor factorization (SFE-NTF). • We decompose each abundance map into a feature layer and a sparse layer. • We design a spatial feature extraction regularization to retrieve the main information of the abundance maps. Estimating endmembers and corresponding abundances from mixed pixels are essential steps for hyperspectral unmixing. In hyperspectral unmixing, obtaining accurate unmixing results is difficult since less prior knowledge is available. Besides, the unmixing results are influenced by noise and highly correlated endmembers, so that the obtained abundance maps exist small values which are not present in the image. In this paper, we separate each abundance map into a feature layer and a sparse layer to protect the obtained abundance maps from the above-mentioned factors. The feature layer represents the main information of the abundance map. And the sparse layer contains outliers dominated by the above-mentioned factors. In particular, we design a feature extraction regularization to describe the feature layer and use a weighted ℓ 1 norm to describe the sparse layer. Then, under the framework of non-negative tensor factorization (NTF), we propose a novel unmixing algorithm named spatial feature extraction NTF (SFE-NTF) for hyperspectral unmixing. The proposed SFE-NTF is based on an augmented multiplicative algorithm. Experimental results on both synthetic and real hyperspectral data demonstrate that the proposed algorithm outperforms other state-of-the-art algorithms. [ABSTRACT FROM AUTHOR]

A method that utilizes two sequential inverse synthetic aperture radar (ISAR) images to estimate the rotation velocity of the target and perform cross-range scaling of ISAR images is proposed. First, features/scatterers that are robust against the change in aspect angle and the scintillation between two sequential ISAR images are extracted. Next, extracted features are matched such that the same feature corresponds between the two sequential ISAR images. To accomplish this, scale-invariant feature transform is adopted for feature extraction, followed by feature matching through nearest-neighbor distance ratio and random sample consensus techniques. In this way, a homography between two sequential ISAR images can be obtained and the rotational velocity of the target can be deduced without prior knowledge of the equivalent rotational center of the target. To demonstrate the effectiveness of the proposed method, the results of a simulation dataset for point scatterers are provided. [ABSTRACT FROM AUTHOR]

One primary challenge in liver segmentation is the fuzzy edge contour. Recently, fully convolutional neural networks (FCNs) have been widely used in liver segmentation because of their adequate feature extraction. Nevertheless, the context among liver slices is still ignored by FCN. To address this issue, we first propose a bidirectional convolutional long short-term memory (BiConvLSTM) to explore contextual information. Meanwhile, the attention gate (AG) is utilized to fuse high-dimensional information from BiConvLSTM to remove irrelevant features. Besides, Shape-Net network is proposed to extend the liver shape pattern by latent space information, which will contribute to reduce the interference of fuzzy boundaries. Finally, the improved active contour loss function with L2 norm acts as a feature constraint. Experimental results on public benchmark datasets show that the proposed method slightly outperforms other newly published methods and achieves good performance for liver segmentation. [ABSTRACT FROM AUTHOR]

Estimation of the sintering state has importance for clinker quality improvements and the safe operation of the rotary kiln. Class imbalanced thermal signals usually pose challenges in feature extraction and abnormal state recognition. In this article, a novel framework that integrates prior knowledge and hidden information is developed for sintering state recognition in the class imbalance condition. For discriminative feature extraction of imbalanced data, a cascaded stack autoencoder (SAE) model is proposed to fuse our prior knowledge and hidden information. The model includes a feature extraction SAE and a deep fusion SAE: the former extracts hidden information from thermal signals, and the latter deeply fuses and compresses our prior knowledge and hidden information. For the class imbalance of sintering samples, we propose a data-dependent kernel modification optimal margin distribution machine (ddKMODM) as a sintering state recognition model. Modifying the original kernel function by a conformal function depending on the data distribution in kernel space, ddKMODM can change the local volume expansion coefficient of the feature space to eliminate the negative effects caused by imbalanced samples. Experiments on real data show that the proposed framework can balance the detection rate of each state in the class imbalanced condition, and its overall sintering state recognition accuracy exceeds 92%. [ABSTRACT FROM AUTHOR]

Conventional networks for object skeleton detection are usually hand-crafted. Despite the effectiveness, hand-crafted network architectures lack the theoretical basis and require intensive prior knowledge to implement representation complementarity for objects/parts in different granularity. In this paper, we propose an adaptive linear span network (AdaLSN), driven by neural architecture search (NAS), to automatically configure and integrate scale-aware features for object skeleton detection. AdaLSN is formulated with the theory of linear span, which provides one of the earliest explanations for multi-scale deep feature fusion. AdaLSN is materialized by defining a mixed unit-pyramid search space, which goes beyond many existing search spaces using unit-level or pyramid-level features. Within the mixed space, we apply genetic architecture search to jointly optimize unit-level operations and pyramid-level connections for adaptive feature space expansion. AdaLSN substantiates its versatility by achieving significantly higher accuracy and latency trade-off compared with the state-of-the-arts. It also demonstrates general applicability to image-to-mask tasks such as edge detection and road extraction. Code is available at https://github.com/sunsmarterjie/SDL-Skeletongithub.com/sunsmarterjie/SDL-Skeleton. [ABSTRACT FROM AUTHOR]

Bearings are important parts in industrial production and are related to the normal operation of mechanical equipment. For bearing fault identification, traditional method often includes feature extraction, which involves professional prior knowledge and is time-consuming. This paper proposes the deep convolution residual network (DCRN) to identify the bearing fault. Based on the end-toend learning characteristics of deep neural networks, this method can directly use raw data for training, and does not require feature extraction. Moreover, under the effect of skip connection, DCRN can exert the powerful fitting ability of deep neural network. In this paper, by stacking residual blocks, three different architecture of DCRN are designed and all three achieve very high accuracy, respectively 99.60%, 99.71% and 99.81%. Compared with other methods, DCRN have better generalization performance. [ABSTRACT FROM AUTHOR]

We present an efficient multi-view stereo (MVS) network for 3D reconstruction from multi-view images. While previous learning based reconstruction approaches performed quite well, most of them estimate depth maps at a fixed resolution using plane sweep volumes with a fixed depth hypothesis at each plane, which requires densely sampled planes for desired accuracy and therefore is difficult to achieve high resolution depth maps. In this paper we introduce a coarse-to-fine depth inference strategy to achieve high resolution depth. This strategy first estimates the depth map at coarsest level, and the depth maps at finer levels are considered as the upsampled depth map from previous level with pixel-wise depth residual. Thus, we narrow the depth searching range with the priori information from previous level and construct new cost volumes from the pixel-wise depth residual to perform depth map refinement. Then the final depth map can be achieved iteratively since all the parameters are shared among different levels. At each level, the self-attention layer is introduced to the feature extraction block for capturing the important information in depth inference task, and the cost volume is generated using similarity measurement instead of the variance based methods used in previous work. Experiments were conducted on three diverse datasets including the DTU benchmark dataset, BlendedMVS dataset and the Tanks and Temples dataset. The results demonstrated that our proposed approach could outperform most state-of-the-arts (SOTA) methods. 2 2 The codebase of this project is at https://github.com/ArthasMil/AACVP-MVSNet. [ABSTRACT FROM AUTHOR]

Collaborative recommender systems are vulnerable to shilling attacks. To address this issue, many methods including supervised and unsupervised have been proposed. However, supervised detection methods require training classifiers and they only apply to detect known types of attacks. The existing unsupervised detection methods need to know the prior knowledge of attacks, otherwise they suffer from low detection precision. In this paper, we present BS-SC, an unsupervised approach for detecting shilling profiles, which does not need to know the attack size or to label the candidate spammers. BS-SC starts from an in-depth analysis of user behaviors and uses two key mechanisms (i.e., behavior features extraction and behavior similarity matrix clustering) to distinguish shilling profiles from genuine ones. The behavior features reflect the behavior difference between genuine and shilling profiles, and the behavior similarity matrix clustering is to cluster shilling profiles based on their highly similar behaviors. Experimental results on the MovieLens and the sampled Amazon review datasets indicate that BS-SC outperforms the baseline unsupervised approaches, even when the prior knowledge is given for them. [ABSTRACT FROM AUTHOR]

Integrating the information of infrared and visible images without human supervision is a long-standing problem. A key technical challenge in this domain is how to extract features from heterogeneous data-sources and fuse them appropriately. Prior deep learning works either extract the middle layers information or use costly training step to improve fusion performance, which limited their performances in cluttered scenes and real-time applications. In this paper, we introduce a novel and pragmatic unsupervised infrared and visible image fusion method based on a pre-trained deep network, which employs a densely connection structure and incorporates the attention mechanism to achieve high fusion performance. Furthermore, we propose to use the cross-dimensional weighting and aggregation to compute the attention map for infrared and visible image fusion. The attention map enables more efficient feature extraction and captures more structure information from source images. We evaluate our method and compare it with ten typical state-of-the-art fusion methods. Extensive experimental results demonstrate that our method achieves state-of-the-art fusion performance in both subjective and objective evaluation. [ABSTRACT FROM AUTHOR]

Existing power analysis techniques rely on strong adversary models with prior knowledge of the leakage or training data. We introduce side-channel analysis with unsupervised learning (SCAUL) that can recover the secret key without requiring prior knowledge or profiling (training). We employ an LSTM auto-encoder to extract features from power traces with high mutual information with the data-dependent samples of the measurements. We demonstrate that by replacing the raw measurements with the auto-encoder features in a classical DPA attack, the efficiency, in terms of required number of measurements for key recovery, improves by 10X. Further, we employ these features to identify a leakage model with sensitivity analysis and multi-layer perceptron (MLP) networks. SCAUL uses the auto-encoder features and the leakage model, obtained in an unsupervised approach, to find the correct key. On a lightweight implementation of AES on Artix-7 FPGA, we show that SCAUL is able to recover the correct key with 3,700 power measurements with random plaintexts, while a DPA attack requires at least 17,400 measurements. Using misaligned traces, with an uncertainty equal to 20 percent of the hardware clock cycle, SCAUL is able to recover the secret key with 12,300 measurements while the DPA attack fails to detect the key. [ABSTRACT FROM AUTHOR]

Reliable detection of anomalies without any prior information is a critical yet challenging task in many applications, not least military and civilian fields. An intelligent anomaly detection system would use the material-specific spectral information in hyperspectral images (HSIs), thereby avoiding the loss of visually confusing objects. However, conventional hyperspectral anomaly detection methods are mainly achieved in an unsupervised way leading to limited performance due to lack of prior knowledge. In this article, we propose a novel autoencoder and adversarial-learning based semisupervised background estimation model (SBEM) that is trained only on the background spectral samples in order to accurately learn the background distribution. In particular, an unsupervised background searching method is firstly conducted on the original HSIs to search the background spectral samples. Our proposed SBEM consists of an encoder, a decoder, and a discriminator to thoroughly capture background distribution. Furthermore, jointly minimizing the reconstruction loss, spectral loss, and adversarial loss during training aids the model to learn the background distribution as required. Experiments on four real HSIs demonstrate that compared to the current state-of-the-art, the proposed framework yields higher detection capability and lower false alarm rate, which shows that it has a significant benefit in the tradeoff between detection accuracy and false alarm rate. [ABSTRACT FROM AUTHOR]

In this paper, we propose a novel method for Multispectral document image binarization (MSdB) through the Non-negative Matrix Factorization (NMF) approach. We propose a three-step MSdB-NMF framework: i) NMF-based feature extraction algorithm by introducing a new optimization problem; ii) post-processing method iii); apply any existing gray/RGB binarization scheme. In the first step, we extract $N$ features out of $B$ spectral bands ($N < B$) and their corresponding coefficient matrix. We introduce a novel objective formulation that considers the robustness (related to the noise and various types of degradations) and sparseness (related to the ratio of text pixels versus the background). We employ the multiplicative updating rules to solve the proposed minimization problem and prove the convergence of the proposed feature extraction algorithm. In the next step, we select an appropriate feature vector, equivalently the corresponding coefficient vector. We propose to select it either visually or automatically via a post-processing method, which uses the benchmark binarization methods as baseline. In the last step, we apply some existing binarization methods such as Sauvola and Howe over the selected coefficient vector. Our proposed binarization framework is applicable for any kind of MS or hyperspectral (HS) document image without considering any prior knowledge such as the side information about the spectral bands of MS/HS document image. We evaluate our proposed binarization framework over two MS document image datasets. The experimental results confirm that our proposed framework outperforms several state-of-the-art binarization schemes including the winner of the contest in MS-TEx-2015. [ABSTRACT FROM AUTHOR]

Predictive maintenance allows industries to keep their production systems available as much as possible. Reducing unforeseen shutdowns to a level that is close to zero has numerous advantages, including production cost savings, a high quality level of both products and processes, and a high safety level. Studies in this field have focused on a novel approach, prognostic health management (PHM), which relies on condition monitoring (CM) for predicting the remaining useful life (RUL) of a system. However, several issues remain in its application to real industrial contexts, e.g., the difficulties in conducting tests simulating each fault condition, the dynamic nature of industrial environments, and the need to handle large amounts of data collected from machinery. In this paper, a data-driven methodology for PHM implementation is proposed, which has the following characteristics: it is unsupervised, i.e., it does not require any prior knowledge regarding fault behaviors and it does not rely on pre-trained classification models, i.e., it can be applied "from scratch"; it can be applied online due to its low computational effort, which makes it suitable for edge computing; and, it includes all of the steps that are involved in a prognostic program, i.e., feature extraction, health indicator (HI) construction, health stage (HS) division, degradation modelling, and RUL prediction. Finally, the proposed methodology is applied in this study to a rotating component. The study results, in terms of the ability of the proposed approach to make a timely prediction of component fault conditions, are promising. [ABSTRACT FROM AUTHOR]

Copyright of Chinese Journal of Liquid Crystal & Displays is the property of Chinese Journal of Liquid Crystal & Displays and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Automated recommendation of machine learning algorithms is receiving a large deal of attention, not only because they can recommend the most suitable algorithms for a new task, but also because they can support efficient hyper-parameter tuning, leading to better machine learning solutions. The automated recommendation can be implemented using meta-learning, learning from previous learning experiences, to create a meta-model able to associate a data set to the predictive performance of machine learning algorithms. Although a large number of publications report the use of meta-learning, reproduction and comparison of meta-learning experiments is a difficult task. The literature lacks extensive and comprehensive public tools that enable the reproducible investigation of the different meta-learning approaches. An alternative to deal with this difficulty is to develop a meta-feature extractor package with the main characterization measures, following uniform guidelines that facilitate the use and inclusion of new meta-features. In this paper, we propose two Meta-Feature Extractor (MFE) packages, written in both Python and R, to fill this lack. The packages follow recent frameworks for meta-feature extraction, aiming to facilitate the reproducibility of meta-learning experiments. [ABSTRACT FROM AUTHOR]

Fine-grained visual categorization is to recognize hundreds of subcategories belonging to the same basic-level category, which is a highly challenging task due to the quite subtle and local visual distinctions among similar subcategories. Most existing methods generally learn part detectors to discover discriminative regions for better categorization performance. However, not all parts are beneficial and indispensable for visual categorization, and the setting of part detector number heavily relies on prior knowledge as well as experimental validation. As is known to all, when we describe the object of an image via textual descriptions, we mainly focus on the pivotal characteristics and rarely pay attention to common characteristics as well as the background areas. This is an involuntary transfer from human visual attention to textual attention, which leads to the fact that textual attention tells us how many and which parts are discriminative and significant to categorization. So, textual attention could help us to discover visual attention in the image. Inspired by this, we propose a fine-grained visual-textual representation learning (VTRL) approach, and its main contributions are: 1) fine-grained visual-textual pattern mining devotes to discovering discriminative visual-textual pairwise information for boosting categorization performance through jointly modeling vision and text with generative adversarial networks, which automatically and adaptively discovers discriminative parts and 2) VTRL jointly combines visual and textual information, which preserves the intra-modality and inter-modality information to generate complementary fine-grained representation, as well as further improves categorization performance. Comprehensive experimental results on the widely used CUB-200-2011 and Oxford Flowers-102 datasets demonstrate the effectiveness of our VTRL approach, which achieves the best categorization accuracy compared with the state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Feature extraction is a major challenge in operational reliability assessment, which requires techniques and prior knowledge. Deep auto-encoder (DAE) is a popular deep learning method and is widely used in feature extraction. However, low generalization ability and structure parameters design are still the major problems of DAE for operational reliability assessment. To overcome the two problems, an ensemble DAE is proposed for operational reliability assessment. Firstly, different structure parameters are employed to design a series of DAEs for feature learning from the measured data. Secondly, a feature ensemble strategy is designed to enhance the generalization ability of the DAE model, in which the features learned by different DAEs are clustered to remove the irrelevant DAEs and select the more general feature subset. Finally, the operational reliability indicator is defined by the Euclidean distance of the selected features and the operational reliability model is developed. The proposed method is utilized to analyze the experimental bearings and the results indicate that the proposed method is effective for operational reliability assessment. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Image & Graphics is the property of Editorial Office of Journal of Image & Graphics and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

• ECSP, ECTP and ECSTP can be used in applications that we have some prior knowledge about the two conditions to be classified. • The performance of the proposed methods ECSP, ECTP, and ECSTP is evaluated on P300 speller data of BCI competitions II and III. In both data sets, our proposed methods significantly outperforms the conventional CSP and CTP methods. • ECSTP reached average character detection accuracy of 98.5% on BCI competition II, which outperforms almost all the other state of the art methods. • An advantage of our proposed methods over many of the other P300 speller classification methods is its significantly less training time (compared to approaches such as eSVM that have an extra channel selection step). Common spatial pattern (CSP) analysis and its extensions have been widely used as feature extraction approaches in the brain-computer interfaces (BCIs). However, most of the CSP-based approaches do not use any prior knowledge that might be available about the two conditions (classes) to be classified. Therefore, their applications are limited to datasets that contain enough variance information about the two conditions. For example, in some event-related potential (ERP) detection applications, such as P300 speller, the information is in the time domain but not in the variance of spatial components. To address this problem, first, we present a novel feature extraction method termed extended common spatial pattern (ECSP) analysis, which uses prior knowledge available from data to produce a broader range of features than that of conventional CSP analysis. Then, similarly, we introduce the extended common temporal pattern (ECTP) analysis. Finally, to exploit both spatial and temporal information, we propose extended common spatial and temporal pattern (ECSTP) analysis. We have used BCI competition III, dataset II as our main dataset to evaluate our proposed methods. In addition, we used two other datasets, namely BCI competition II, dataset IIb and BCI competition IV, dataset IIb, to further evaluate the performance of the proposed methods. In All the datasets, the proposed methods significantly outperform the conventional CSP, CTP, and CSTP methods. More specifically, ECSTP has the best performance among the proposed methods. Moreover, classification results show that the proposed methods are competitive with other state of the art methods applied to these datasets. [ABSTRACT FROM AUTHOR]

Copyright of Transactions of Beijing Institute of Technology is the property of Beijing University of Technology and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Faulted line detection is a key step of intelligent fault diagnosis of distribution systems, laying the foundation for the further fault location and service restoration. A novel single line-to-ground (SLG) faulted line detection method based on the feature fusion framework is proposed. In the proposed framework, one-dimensional convolutional neural network is employed as a powerful tool to extract more effective features. In addition, there is an imbalance phenomenon between data of the faulted line and healthy lines when a data-driven model is used in the faulted line detection. The proposed framework offers an avenue for overcoming it and improves the accuracy of detection. Considering the limited data of SLG faults in actual power systems, prior knowledge of SLG fault detection is integrated into the data-driven model, which proves useful in reducing dependence on the training data quantity. The experiments verified the superior performance of the proposed feature fusion framework-based method. [ABSTRACT FROM AUTHOR]

Dental hard tissue registration is very challenging due to the complexity of histological structure, the variability of imaging modality and image quality. To benefit from the advantages of the relative works, this paper proposed a novel dental hard tissue auto-registration method based on multi-modality dental images. It employed the modified ICP algorithm and was especially effective for dental fluorescence imaging and the reflectance imaging. The algorithm firstly pre-processed the dental images to decrease the impact of imbalance local illumination. Secondly, it investigated a robust feature points extraction strategy, i. e. calculated the boundaries of hard tissue as initial feature point sets, and further extracted the lesion areas based on the prior knowledge of oral histopathology for the refinement of feature points. Finally, this paper executed the modified ICP algorithm on the basis of the ICP iterative method and robust loss function to register dental image. The experimental results show that the proposed method is not only robust, but also can converge more quickly and achieve higher accuracy. [ABSTRACT FROM AUTHOR]

Current learning-based dehazing methods simply rely on the paired synthetic datasets and physical models, which can hardly describe the complicated degradation for practical applications. These methods still struggle to achieve haze removal, color distortion, and detail restoration instantly, and they ignore the frequency characteristic differences and prior knowledge importance. To address these problems, we propose an unpaired stage-wise framework integrating frequency-guided filtering and progressive physics learning in an adversarial dehazing network, called FPD-Net. To be specific, a guided filter based on frequency information is employed to decompose the high and low frequency components for better feature extraction. We further merge the prior and physical knowledge to form progressive physics learning, which produces pleasing haze-free outputs with high visibility and reality. For better atmospheric light estimation, the variational auto-encoder and Kullback–Leibler loss are included to represent the illumination message. Extensive experiments on both synthetic and real datasets prove that our designed FPD-Net achieves better performance visually and quantitatively than the comparison dehazing models. [ABSTRACT FROM AUTHOR]

In recent years, the rapid development of deep learning has led to an extensive amount of research into machining condition monitoring technologies. However, the complexity of multi-modal signals present in actual machining scenarios poses a significant challenge to signal fusion, and most models find it difficult to balance the relationship between performance and deployment cost. In this research, a multi-scale hybrid attention aggregation network (MHAAN) is proposed for multi-modal monitoring in laser-induced thermal-crack processing. Based on the feature distribution of visual signals and acoustic emission (AE) signals, the multi spatial scale aggregation module (MSSAM) and Multi frequency domain aggregation module (MFDAM) are designed. A cross-modal dot-product interaction module is employed to facilitate multilevel interactive fusion of the extracted feature vectors. This module enables feature transfer across different modalities, and then extracts the multi-model fusion feature. The practicality of MHAAN has been validated using a dataset obtained from the process of glass cutting with laser-induced thermal-crack propagation (LITP). Compared with commonly used models, the superiority of MHAAN in multi-modal feature extraction and fusion is demonstrated. Finally, the effectiveness of the module design based on prior knowledge is proven through ablation experiments, improving the interpretability of the model. [ABSTRACT FROM AUTHOR]