Unsupervised visible-infrared person re-identification (USVI-ReID) aims to match individuals across visible and infrared cameras without relying on any annotation. Given the significant gap across visible and infrared modality, estimating reliable cross-modality association becomes a major challenge in USVI-ReID. Existing methods usually adopt optimal transport to associate the intra-modality clusters, which is prone to propagating the local cluster errors, and also overlooks global instance-level relations. By mining and attending to the visible-infrared modality bias, this paper focuses on addressing cross-modality learning from two aspects: bias-mitigated global association and modality-invariant representation learning. Motivated by the camera-aware distance rectification in single-modality re-ID, we propose modality-aware Jaccard distance to mitigate the distance bias caused by modality discrepancy, so that more reliable cross-modality associations can be estimated through global clustering. To further improve cross-modality representation learning, a `split-and-contrast' strategy is designed to obtain modality-specific global prototypes. By explicitly aligning these prototypes under global association guidance, modality-invariant yet ID-discriminative representation learning can be achieved. While conceptually simple, our method obtains state-of-the-art performance on benchmark VI-ReID datasets and outperforms existing methods by a significant margin, validating its effectiveness.

A converging pattern represents the process in which a collection of moving objects gradually converges toward a target area from various directions and eventually forms a dense group. Unlike most existing group patterns, it indicates the early formation of group events, which has a significant application for predicting and detecting emergency events. Existing studies of the converging pattern merely discover patterns from historical trajectories in an offline manner. However, online mining over streaming trajectories has a more practical impact in some real-world scenarios like real-time traffic monitoring . In this paper, we investigate online algorithms that enable converging pattern mining over network-constrained streaming trajectories of moving objects. To achieve synchronization with the speed of trajectory updates, we propose an incremental density-based clustering algorithm in the road network called I D C R N and a converging monitoring method to detect converging patterns in real-time. To efficiently retrieve the constantly evolving spatial relationship among objects in road networks with a large search space and an intractable computation complexity for network distance, we propose a dual index called M O R N to support continuous neighborhood query and cluster pruning in the road network . Extensive experiments with real and synthetic datasets validate the efficiency of our proposed index and methods.

Interactive scenario understanding in traffic has been widely studied for autonomous driving and accident analysis. However, existing studies focus on the interactive behaviors of pedestrians and vehicles from an independent perspective, overlooking the intricate dynamics of their interactions, particularly in the context of violations where vehicles fail to yield to pedestrians. To fill this gap, we introduceCrosswalk, a novel dataset with over 7.7k annotated events from 10 hours of surveillance video, providing an integrated view of pedestrian-vehicle interactions in both violation and non-violation scenarios.Crosswalkhighlights the interdependencies between traffic participants and offers event-level evaluation metrics for scene understanding. Besides,Crosswalkposes two significant challenges in the interactive scenario understanding task, which are capturing subtle patterns of violations and comprehending intricate contextual interactions. To this end, we also propose a multiscale aggregation network that detects violation patterns by integrating temporal and spatial representations while employing hierarchical spatiotemporal fusion. Moreover, we explore state-of-the-art video understanding methods with an event awareness framework as a baseline.Crosswalkaims to advance the comprehensive understanding of interactive scenarios and specific events in real-world settings. The dataset and code are available at: https://github.com/Nanasaki-Ai/Crosswalk

Most existing image quality assessment methods need to be retrained when dealing with a new type of task. This approach wastes computing resources and time. Therefore, these methods fail to suit the application scenarios that require processing of multi-type image quality assessment tasks. In the human visual system, the eyes of human tend to pay varying degrees of attention to different regions. Inspired by this system, this paper proposes a multi-type image quality assessment method based on multi-region deep feature fusion under meta-learning (MMQA). First, we utilize the differences in the structural information to screen out salient and non-salient regions. Second, a deep multi-stream network is designed to comprehensively consider and fuse different features related to the quality in salient regions, non-salient regions and the entire image. Third, meta-learning is applied to quickly learn and update the parameters of the model when facing new types of images. By summarizing the prior knowledge in the training of one type of task, the model can be quickly fine-tuned for other types of images. The experimental results demonstrate that the proposed method has advantages over the existing methods in generalization and robustness. Furthermore, the proposed method can adapt well to different distortion types and different image types quickly and accurately.

Label distribution learning has gained an increasing amount of attention in comparison to single-label and multi-label learning due to its more universal capacity to communicate label ambiguity. Unfortunately, label distribution learning cannot be used directly in many real tasks, because it is very difficult to obtain the label distribution datasets, and many training sets only contain simple logical labels. To resolve this problem and recover the label distributions from the logical labels, label enhancement is proposed. This paper proposes a novel label enhancement algorithm called Privileged Label Enhancement with Adaptive Graph(PLEAG). PLEAG first apply adaptive graph to capture the hidden information between instances and treat it as privileged information. As a result, the similarity matrix of instances is not only influenced by the feature space, but is also adaptively modified in accordance with the degree of similarity between instances in the label space. Then, we adopt RSVM+ model in the paradigm of LUPI (learning with privileged information) to handle the new dataset with privileged information in order to gain better learning effect. Our comparison experiments on 12 datasets show that our proposed algorithm PLEAG , is more accurate than prior label enhancement algorithms for recovering label distribution from logical labels.

Abstract In this article, we propose a new multi‐label enhancement manifold learning algorithm to solve the vehicle video classification problem. Predicting multiple objects in a traffic video image is a challenging problem. Traditional multi‐label classification methods can solve the problem of simultaneous detection of multiple labels, but cannot handle high‐dimensional streaming video data. Our idea is to use label distribution learning (LDL) to enrich the label space and improve label recognition in the original label space. We use the feature function representing the manifold structure to guide the geometric meaning of the label space and transform the local topology from the feature space to the label space. We first build a label distribution learner. Next, use the LDL model for classification. The similarity between the two distributions is measured by Bayesian divergence, and the label distribution is learned through the maximum entropy model and the objective function of this article is established. Finally, an enhanced label model of the manifold space is established to reduce the dimensionality of the feature matrix generated during the training phase, so that the supervised information in the label manifold can be used in the incremental manifold space to improve the accuracy of feature extraction. Compared to the latest multi‐label learning methods, our multi‐label enhancement manifold learning method has advantages in predicting performance.

Significant development of video person re-identification has been witnessed in recent years with deep learning technologies. Due to the complexity of human pose changes and the similarity between different individuals, learning discriminative features is still a challenging part of the video person re-identification task. To get rid of the effects of pose misalignment while keep the similarity of human appearance, in this paper, we propose a Pose and Similarity based Graph Neural Network in a decoupled manner, which consists of three independent branches to emphasize the respective roles of pose, local similarity and global similarity in the final descriptions. Compared to traditional Convolutional Neural Networks which tend to output similar global features in the case of highly similar pedestrians, the developed Graph Neural Networks are able to explore local semantic relationships between body parts, resulting in more discriminative features. To further eliminate the pose variation, we incorporate human skeleton information for feature map segmentation. Specifically, we propose to take a tree structure as the pose-aware adjacency graph of blocks in a person frame, which reveals the inherent connections within a human body. Experimental results on four widely used datasets demonstrate the effectiveness of our method.

Real-world multilabel data are high dimensional, and directly using them for label distribution learning (LDL) will incur extensive computational costs. We propose a multilabel distribution learning algorithm based on multioutput regression through manifold learning, referred to as MDLRML. By exploiting smooth, similar spaces' information provided by the samples' manifold learning and LDL, we link the two spaces' manifolds. This facilitates using the topological relationship of the manifolds in the feature space to guide the manifold construction of the label space. The smoothest regression function is used to fit the manifold data, and a locally constrained multioutput regression is designed to improve the data's local fitting. Based on the regression results, we enhance the logical labels into the label distributions, thereby mining and revealing the label's hidden information regarding importance or significance. Extensive experimental results using real-world multilabel datasets show that the proposed MDLRML algorithm significantly improves the multilabel distribution learning accuracy and efficiency over several existing state-of-the-art schemes.

We propose a novel probabilistic label enhancement algorithm, called PLEA, to solve challenging label distribution learning (LDL) for multi-label classification problems. We adopt the well-known maximum entropy model based label distribution learner. However, unlike the existing LDL algorithms based on the maximum entropy model, we propose to use manifold learning to enhance the label distribution learner. Specifically, the supervised information in the label manifold is utilized in the feature manifold space construction to improve the accuracy of feature extraction, while dramatically reducing the feature dimension. Then the robust linear regression is employed to estimate the label distributions associated with the extracted reduced-dimension features. Using the enhanced reduced-dimension features and their associated estimated label distributions in the maximum entropy model, the unknown true label distributions can be estimated more accurately, while imposing considerably lower computational complexity. We evaluate the proposed PLEA method on a wide-range artificial and high-dimensional real-world datasets. Experimental results obtained demonstrate that our proposed PLEA method has advantages in LDL accuracy and runtime performance, compared to the latest multi-label LDL approaches. The results also show that our PLEA compares favourably with the state-of-the-arts multi-label learning algorithms for classification tasks.

MOTIVATION: Narrative visualization for scientific data explorations can help users better understand the domain knowledge, because narrative visualizations often present a sequence of facts and observations linked together by a unifying theme or argument. Narrative visualization in immersive environments can provide users with an intuitive experience to interactively explore the scientific data, because immersive environments provide a brand new strategy for interactive scientific data visualization and exploration. However, it is challenging to develop narrative scientific visualization in immersive environments. In this paper, we propose an immersive narrative visualization tool to create and customize scientific data explorations for ordinary users with little knowledge about programming on scientific visualization, They are allowed to define POIs (point of interests) conveniently by the handler of an immersive device. RESULTS: Automatic exploration animations with narrative annotations can be generated by the gradual transitions between consecutive POI pairs. Besides, interactive slicing can be also controlled by device handler. Evaluations including user study and case study are designed and conducted to show the usability and effectiveness of the proposed tool. AVAILABILITY: Related information can be accessed at: https://dabigtou.github.io/richenliu/.

With the fast development of various computing paradigms, the amount of data is rapidly increasing that brings the huge storage overhead. However, the existing data deduplication techniques do not make full use of similarity detection to improve the storage efficiency and data transmission rate. In this paper, we study the problem of utilising the duplicate and resemblance detection techniques to further compress data. We first present a framework of FIDCS-ERD, a feature-based intelligent deduplication compression system with extreme resemblance detection. We also introduce the main components and the detailed workflow of our compression system. We propose a content-defined chunking algorithm for duplicate detection and a Bloom filter-based resemblance detection algorithm. FIDCS-ERD implements the intelligent file chunking and the fast duplicate and resemblance detection. By extensive experiments over the real datasets, we demonstrate that FIDCS-ERD has better compression effect and more accurate resemblance detection compared to the existing approaches.

Group pattern mining from spatio-temporal trajectories of moving objects have gained significant attentions due to the prevalence of location-acquisition devices and tracking technologies. In this work, we propose a new group pattern, named converging, which is a group of moving objects that converge from different directions for a certain time period. Examples of convergings may include traffic jams, troop assembly, serious stampedes, and other public congregations. As a proof-of-concept, we implemented a visual analytic system CPM based on road-network constrained trajectories to detect converging events in road networks. A user-friendly interface is designed to help users gain insights into converging events from spatial and temporal aspects. Finally, we demonstrate the effectiveness and efficiency of our system by using a real dataset.

Image quality assessment (IQA) is a meaningful research topic to meet the increasing demand of high‐quality image. The degradation of image quality will cause changes in image structural information. Meanwhile, human visual system is sensitive to changes in structural information. This finding motivates us to utilise structural information for proposing IQA method which is consistent with human visual perception. Recently, IQA methods are mainly focused on individual image type, e.g. natural image or screen content image (SCI), thus, the authors proposed a novel no‐reference IQA method which can be suitable for both natural image and SCI. The proposed method is based on structural information analysis. For each image, they first obtain the grey‐scale fluctuation maps (GFMs) in four detection directions. After that, the grey‐scale fluctuation direction map (GFD) of certain image can be acquired via its GFMs. Based on the GFMs and GFD, the structural features of each image are extracted, and then collected and transformed to feature vectors. Subsequently, the IQA model is trained by support vector regression. The experimental results on the public databases demonstrate the proposed method can predict image quality accurately for both natural image and SCI, and the performance is competitive with prevalent methods.

Summary Underground flow path (UFP) is one of the most significant stratigraphic structures in revealing the distribution of oil or gas from seismic data. We design a domain‐specific visualization system to extract the stratigraphic structures by seed point tracing and explore the seismic data by graph interactions. The seeds are automatically generated by kernel function–based density gradients computation. Users are allowed to adjust the recommended seeds by fine‐tuning them with visual interactions. The seeds are further merged by a weighted quick‐union algorithm to get the link information to construct a graph. Different types of nodes in the graph are designed to enable users to explore the extracted UFP structures intuitively. Finally, we evaluated the proposed approach by performance tests, sensitivity tests, and ground truth tests. The feedback from the domain experts demonstrates that the proposed visualization tool improved the capability of revealing the distribution and geostructures of UFPs compared with the existing methods.

The widespread use of numerical simulations in different scientific domains provides a variety of research opportunities. They often output a great deal of spatio-temporal simulation data, which are traditionally characterized as single-run, multi-run, multi-variate, multi-modal and multi-dimensional. From the perspective of data exploration and analysis, we noticed that many works focusing on spatio-temporal simulation data often share similar exploration techniques, for example, the exploration schemes designed in simulation space, parameter space, feature space and combinations of them. However, it lacks a survey to have a systematic overview of the essential commonalities shared by those works. In this survey, we take a novel multi-space perspective to categorize the state-of-the-art works into three major categories. Specifically, the works are characterized as using similar techniques such as visual designs in simulation space (e.g, visual mapping, boxplot-based visual summarization, etc.), parameter space analysis (e.g, visual steering, parameter space projection, etc.) and data processing in feature space (e.g, feature definition and extraction, sampling, reduction and clustering of simulation data, etc.).

Summary In big data era, how to process large‐scale data stream is one of the existing challenges. Feature extraction method has attracted much attention because of its effectiveness to data classification. Traditional classification algorithms may take less advantage of labeled samples information. Online learning and out‐of‐sample problems are also hot topics recently. To solve these problems, a novel algorithm namedsemi‐supervisedincrementalfeatureextraction algorithm is proposed in this paper. First, we extract feature incrementally in unsupervised way. Then we propose a semi‐supervised subspace learning algorithm by taking advantage of class information to adjustk‐nearest neighbor weights. Third, we combine the unsupervised and semi‐supervised feature extraction approaches to obtain objective function, in order to solve the out‐of‐sample learning problem. Experiments have been carried out on Machine learning datasets of University of California Irvine (UCI) datasets and real‐world face image datasets (Olivetti faces (ORL), Yale, YaleB, and Rendered face). To demonstrate the proposed algorithm's expandability to process the large‐scale data stream, classification experiments using Spark skill in parallel computation environment are performed, with comparisons with some related semi‐supervised feature extraction methods. The experiment results and computational complex comparison demonstrate that the proposed algorithm can obtain good performance. Copyright © 2016 John Wiley & Sons, Ltd.

Summary In the field of machine learning, feature extraction is one of the most important preprocessing in data classification for its effectiveness, and now it has attracted much extensive attention for large‐scale data stream preprocessing step, especially in the era of big data. Motivated by the advantages of unsupervised and supervised feature extraction, which are two desirable and promising characteristics for dimension reduction, a new semisupervised local preserving embedding algorithm based on maximum margin criterion (SLPE/MMC) is proposed in this paper. First, the objective functions of maximum margin criterion (MMC) and neighborhood preserving embedding (NPE) are combined to get the first objective function of SLPE/MMC. Then, in order to overcome the out‐of‐sample problem, a linear transformation is introduced to construct the second objective function. At last, the whole optimal objective function is constructed by combing the two objective functions together. The proposed algorithm has effectively taken advantage of the sample's supervised information and keeps the geometry structure and the class discrimination information of the manifold. Experiments on face datasets Yale, CMU PIE, and AR datasets are performed to evaluate the classification accuracy of SLPE/MMC. The experimental results and time complexity comparisons have demonstrated the effectiveness of the proposed method.

Abstract Accurate fruit classification is difficult to accomplish because of the similarities among the various categories. In this paper, we proposed a novel fruit‐classification system, with the goal of recognizing fruits in a more efficient way. Our methodology included the following steps. First, a four‐step pre‐processing was employed. Second, the features (colour, shape, and texture) were extracted. Third, we utilized principal component analysis to remove excessive features. Fourth, a novel fruit‐classification system based on biogeography‐based optimization (BBO) and feedforward neural network (FNN) was proposed, with the short name of BBO‐FNN. The experiment employed over 1653 chromatic fruit images (18 categories) by fivefold stratified cross‐validation. The results showed that the proposed BBO‐FNN yielded an overall accuracy of 89.11%, which was higher than the five state‐of‐the‐art methods: genetic algorithm‐FNN, artificial bee colony‐FNN, particle swarm optimization‐FNN, kernel support vector machine, and ant colony optimization‐FNN. Also, the BBO‐FNN achieved the same accuracy as fitness‐scaling chaotic artificial bee colony‐FNN, but it performed much faster than the latter. The proposed BBO‐FNN was effective in fruit‐classification in terms of classification accuracy and computation time. This indicated that it can be applied in credible use.