Search Results (14)

Effectively classifying each pixel in an image is an important research topic in semantic segmentation. The Existing methods typically require the network to directly generate a feature map of the same size as the original image and classify each pixel, which makes it difficult for the network to fully leverage the representations from the backbone. To handle this challenge, this paper proposes a method named mask classification combined with an adaptive dilated convolution network (MCDCNet). Firstly, a Vision Transformer (ViT)-based module is employed to capture contextual features as the backbone. Secondly, the Spatial Extraction Module (SEM) is proposed to extract multi-scale spatial information through adaptive dilated convolution while preserving the original feature size. This spatial information is then integrated into the corresponding contextual features to enhance the representation. Finally, a novel inference process is proposed that incorporates the instance activation map (IAM)-based decoder for semantic segmentation, thereby enhancing the network’s capability to capture and comprehend semantic features. The experimental results demonstrate that our network significantly outperforms other per-pixel classification networks across several semantic segmentation datasets. In particular, on Cityscapes, MCDCNet achieves 80.3 mIoU with 11.8 M Params, demonstrating that the network is able to deliver a strong segmentation performance while maintaining a relatively low parameter count. Full article

Infrared-visible image fusion (IVIF) is an important part of multimodal image fusion (MMF). Our goal is to combine useful information from infrared and visible sources to produce strong, detailed, fused images that help people understand scenes better. However, most existing fusion methods based on convolutional neural networks extract cross-modal local features without fully utilizing long-range contextual information. This limitation reduces performance, especially in complex scenarios. To address this issue, we propose TCTFusion, a three-branch cross-modal transformer for visible–infrared image fusion. The model includes a shallow feature module (SFM), a frequency decomposition module (FDM), and an information aggregation module (IAM). The three branches specifically receive input from infrared, visible, and concatenated images. The SFM extracts cross-modal shallow features using residual connections with shared weights. The FDM then captures low-frequency global information across modalities and high-frequency local information within each modality. The IAM aggregates complementary cross-modal features, enabling the full interaction between different modalities. Finally, the decoder generates the fused image. Additionally, we introduce pixel loss and structural loss to significantly improve the model’s overall performance. Extensive experiments on mainstream datasets demonstrate that TCTFusion outperforms other state-of-the-art methods in both qualitative and quantitative evaluations. Full article

While a flurry of studies and Integrated Assessment Models (IAMs) have independently investigated the impacts of switching mitigation policies in response to different climate scenarios, little is understood about the feedback effect of how human risk perceptions of climate change could contribute to switching climate mitigation policies. This study presents a novel machine learning approach, utilizing a probabilistic structural equation model (PSEM), for understanding complex interactions among climate risk perceptions, beliefs about climate science, political ideology, demographic factors, and their combined effects on support for mitigation policies. We use machine learning-based PSEM to identify the latent variables and quantify their complex interaction effects on support for climate policy. As opposed to a priori clustering of manifest variables into latent variables that is implemented in traditional SEMs, the novel PSEM presented in this study uses unsupervised algorithms to identify data-driven clustering of manifest variables into latent variables. Further, information theoretic metrics are used to estimate both the structural relationships among latent variables and the optimal number of classes within each latent variable. The PSEM yields an R² of 92.2% derived from the “Climate Change in the American Mind” dataset (2008–2018 [N = 22,416]), which is a substantial improvement over a traditional regression analysis-based study applied to the CCAM dataset that identified five manifest variables to account for 51% of the variance in policy support. The PSEM uncovers a previously unidentified class of “lukewarm supporters” (~59% of the US population), different from strong supporters (27%) and opposers (13%). These lukewarm supporters represent a wide swath of the US population, but their support may be capricious and sensitive to the details of the policy and how it is implemented. Individual survey items clustered into latent variables reveal that the public does not respond to “climate risk perceptions” as a single construct in their minds. Instead, PSEM path analysis supports dual processing theory: analytical and affective (emotional) risk perceptions are identified as separate, unique factors, which, along with climate beliefs, political ideology, and race, explain much of the variability in the American public’s support for climate policy. The machine learning approach demonstrates that complex interaction effects of belief states combined with analytical and affective risk perceptions; as well as political ideology, party, and race, will need to be considered for informing the design of feedback loops in IAMs that endogenously feedback the impacts of global climate change on the evolution of climate mitigation policies. Full article

Hydraulic systems are critical components of mechanical equipment, and effective fault diagnosis is essential for minimizing maintenance costs and enhancing system reliability. In practical applications, data from hydraulic systems are collected with varying sampling frequencies, coupled with complex interdependencies within the data, which poses challenges for existing fault diagnosis algorithms. To solve the above problems, this paper proposes an intelligent fault diagnosis of a hydraulic system based on a multiscale one-dimensional convolution neural network with a multiattention mechanism (MA-MS1DCNN). The proposed method first extracts features from multirate data samples using a parallel 1DCNN with different receptive fields. Next, a Hybrid Attention Module (HAM) is proposed, consisting of two submodules: the Correlation Attention Module (CAM) and the Importance Attention Module (IAM), which aim to meticulously and comprehensively model the complex relationships between channel features. Subsequently, to effectively utilize the feature information of different frequencies, the HAM is integrated into the 1DCNN to form the MA-MS1DCNN. Finally, the proposed method is evaluated and experimentally compared using the UCI hydraulic system dataset. The results demonstrate that, compared to existing methods such as Shapelet, MCIFM, and CNNs, the proposed method shows superior diagnostic performance. Full article

In the realm of offline handwritten text recognition, numerous normalization algorithms have been developed over the years to serve as preprocessing steps prior to applying automatic recognition models to handwritten text scanned images. These algorithms have demonstrated effectiveness in enhancing the overall performance of recognition architectures. However, many of these methods rely heavily on heuristic strategies that are not seamlessly integrated with the recognition architecture itself. This paper introduces the use of a Pix2Pix trainable model, a specific type of conditional generative adversarial network, as the method to normalize handwritten text images. Also, this algorithm can be seamlessly integrated as the initial stage of any deep learning architecture designed for handwritten recognition tasks. All of this facilitates training the normalization and recognition components as a unified whole, while still maintaining some interpretability of each module. Our proposed normalization approach learns from a blend of heuristic transformations applied to text images, aiming to mitigate the impact of intra-personal handwriting variability among different writers. As a result, it achieves slope and slant normalizations, alongside other conventional preprocessing objectives, such as normalizing the size of text ascenders and descenders. We will demonstrate that the proposed architecture replicates, and in certain cases surpasses, the results of a widely used heuristic algorithm across two metrics and when integrated as the first step of a deep recognition architecture. Full article

An Integrated Assessment Model (IAM) was employed to develop a Narrative Policy Framework (NPF) and a quantitative model to investigate the changes in land use within the Brazilian Amazon. The process began by creating a theoretical NPF using a ‘systems thinking’ approach. Subsequently, a ‘system dynamic model’ was built based on an extensive review of the literature and on multiple quantitative datasets to simulate the impacts of the NPF, specifically focusing on the conversion of forests into open land for ranching and the implementation of soil management practices as a macro-level policy aimed at preserving soil quality and ranching yields. Various fallow scenarios were tested to simulate their effects on deforestation patterns. The results indicate that implementing fallow practices as a policy measure could reduce deforestation rates while simultaneously ensuring sustainable long-term agricultural productivity, thus diminishing the necessity to clear new forest land. Moreover, when combined with payments for avoided deforestation, such as REDD+ carbon offsets, the opportunity costs associated with ranching land can be utilized to compensate for the loss of gross income resulting from the policy. A sensitivity analysis was conducted to assess the significance of different model variables, revealing that lower cattle prices require resources for REDD+ payments, and vice-versa. The findings indicate that, at the macro level, payments between USD 2.5 and USD 5.0 per MgC ha⁻¹ have the potential to compensate the foregone cattle production from not converting forest into ranching land. This study demonstrates that employing an IAM with a systems approach facilitates the participation of various stakeholders, including farmers and landowners, in policy discussions. It also enables the establishment of effective land use and management policies that mitigate deforestation and soil degradation, making it a robust initiative to address environmental, climate change, and economic sustainability issues. Full article

Convolutional neural networks (CNNs) have demonstrated impressive performance and have been broadly applied in hyperspectral image (HSI) classification. However, two challenging problems still exist: the first challenge is that redundant information is averse to feature learning, which damages the classification performance; the second challenge is that most of the existing classification methods only focus on single-scale feature extraction, resulting in underutilization of information. To resolve the two preceding issues, this article proposes a multiscale cross interaction attention network (MCIANet) for HSI classification. First, an interaction attention module (IAM) is designed to highlight the distinguishability of HSI and dispel redundant information. Then, a multiscale cross feature extraction module (MCFEM) is constructed to detect spectral–spatial features at different scales, convolutional layers, and branches, which can increase the diversity of spectral–spatial features. Finally, we introduce global average pooling to compress multiscale spectral–spatial features and utilize two fully connection layers, two dropout layers to obtain the output classification results. Massive experiments on three benchmark datasets demonstrate the superiority of our presented method compared with the state-of-the-art methods. Full article

Writer recognition based on a small amount of handwritten text is one of the most challenging deep learning problems because of the implicit characteristics of handwriting styles. In a deep convolutional neural network, writer recognition based on supervised learning has shown great success. These supervised methods typically require a lot of annotated data. However, collecting annotated data is expensive. Although unsupervised writer recognition methods may address data annotation issues significantly, they often fail to capture sufficient feature relationships and usually perform less efficiently than supervised learning methods. Self-supervised learning may solve the unlabeled dataset issue and train the unsupervised datasets in a supervised manner. This paper introduces Self-Writer, a self-supervised writer recognition approach dealing with unlabeled data. The proposed scheme generates clusterable embeddings from a small fixed-length image frame such as a text block. The training strategy presumes that a small image frame of handwritten text should include the writer’s handwriting characteristics. We construct pairwise constraints and nongenerative augmentation to train Siamese architecture to generate embeddings depending on such an assumption. Self-Writer is evaluated on the two most widely used datasets, IAM and CVL, on pairwise and triplet architecture. We find Self-Writer to be convincing in achieving satisfactory performance using pairwise architectures. Full article

At the crossroads of sociology and international relations, this interdisciplinary and comparative research article explores how the COVID-19 outbreak has impacted China–Europe relations. Unfolding the critical moments of the COVID-19 outbreak, this article characterizes the evolution of China–Europe relations with regard to the facemask. This simple object of self-protection against the coronavirus strikingly became a source of contention between peoples and states. In the face of this situation, we argue that the facemask is the prism through which to illustrate (1) the transnational links between China and its overseas population, (2) the changing social perceptions of China and Chinese-looking people in European societies, and (3) the advent of China’s health diplomacy and its reception in Europe. Comparing two European settings—France and the United Kingdom (UK)—the common denominator appears to be the reduced trust, if not outright distrust, between individuals and communities in the French and British contexts, and in Sino–French and Sino–British relations at the transnational level. Combining critical juncture theory and (dis)trust in international relations as our analytical framework, this article examines how the facemask became a politicized object, both between states and between Mainland China and its overseas population, as the epidemic unfolded throughout Europe. Adopting a qualitative approach, our dataset comprises the analysis of official speeches and statements; press releases; traditional and social media content (especially through hashtags such as #JeNeSuisPasUnVirus, #IAmNotAVirus, #CoronaRacism, etc.); and interviews with Chinese, French, and British community members. Full article

Internet of Things (IoT) environments produce large amounts of data that are challenging to analyze. The most challenging aspect is reducing the quantity of consumed resources and time required to retrain a machine learning model as new data records arrive. Therefore, for big data analytics in IoT environments where datasets are highly dynamic, evolving over time, it is highly advised to adopt an online (also called incremental) machine learning model that can analyze incoming data instantaneously, rather than an offline model (also called static), that should be retrained on the entire dataset as new records arrive. The main contribution of this paper is to introduce the Incremental Ant-Miner (IAM), a machine learning algorithm for online prediction based on one of the most well-established machine learning algorithms, Ant-Miner. IAM classifier tackles the challenge of reducing the time and space overheads associated with the classic offline classifiers, when used for online prediction. IAM can be exploited in managing dynamic environments to ensure timely and space-efficient prediction, achieving high accuracy, precision, recall, and F-measure scores. To show its effectiveness, the proposed IAM was run on six different datasets from different domains, namely horse colic, credit cards, flags, ionosphere, and two breast cancer datasets. The performance of the proposed model was compared to ten state-of-the-art classifiers: naive Bayes, logistic regression, multilayer perceptron, support vector machine, K*, adaptive boosting (AdaBoost), bagging, Projective Adaptive Resonance Theory (PART), decision tree (C4.5), and random forest. The experimental results illustrate the superiority of IAM as it outperformed all the benchmarks in nearly all performance measures. Additionally, IAM only needs to be rerun on the new data increment rather than the entire big dataset on the arrival of new data records, which makes IAM better in time- and resource-saving. These results demonstrate the strong potential and efficiency of the IAM classifier for big data analytics in various areas. Full article

Offline handwritten text recognition (HTR) for historical documents aims for effective transcription by addressing challenges that originate from the low quality of manuscripts under study as well as from several particularities which are related to the historical period of writing. In this paper, the challenge in HTR is related to a focused goal of the transcription of Greek historical manuscripts that contain several particularities. To this end, in this paper, a convolutional recurrent neural network architecture is proposed that comprises octave convolution and recurrent units which use effective gated mechanisms. The proposed architecture has been evaluated on three newly created collections from Greek historical handwritten documents that will be made publicly available for research purposes as well as on standard datasets like IAM and RIMES. For evaluation we perform a concise study which shows that compared to state of the art architectures, the proposed one deals effectively with the challenging Greek historical manuscripts. Full article

Handwriting recognition refers to recognizing a handwritten input that includes character(s) or digit(s) based on an image. Because most applications of handwriting recognition in real life contain sequential text in various languages, there is a need to develop a dynamic handwriting recognition system. Inspired by the neuroevolutionary technique, this paper proposes a Dynamically Configurable Convolutional Recurrent Neural Network (DC-CRNN) for the handwriting recognition sequence modeling task. The proposed DC-CRNN is based on the Salp Swarm Optimization Algorithm (SSA), which generates the optimal structure and hyperparameters for Convolutional Recurrent Neural Networks (CRNNs). In addition, we investigate two types of encoding techniques used to translate the output of optimization to a CRNN recognizer. Finally, we proposed a novel hybridized SSA with Late Acceptance Hill-Climbing (LAHC) to improve the exploitation process. We conducted our experiments on two well-known datasets, IAM and IFN/ENIT, which include both the Arabic and English languages. The experimental results have shown that LAHC significantly improves the SSA search process. Therefore, the proposed DC-CRNN outperforms the handcrafted CRNN methods. Full article

Handwritten keyword spotting (KWS) is of great interest to the document image research community. In this work, we propose a learning-free keyword spotting method following query by example (QBE) setting for handwritten documents. It consists of four key processes: pre-processing, vertical zone division, feature extraction, and feature matching. The pre-processing step deals with the noise found in the word images, and the skewness of the handwritings caused by the varied writing styles of the individuals. Next, the vertical zone division splits the word image into several zones. The number of vertical zones is guided by the number of letters in the query word image. To obtain this information (i.e., number of letters in a query word image) during experimentation, we use the text encoding of the query word image. The user provides the information to the system. The feature extraction process involves the use of the Hough transform. The last step is feature matching, which first compares the features extracted from the word images and then generates a similarity score. The performance of this algorithm has been tested on three publicly available datasets: IAM, QUWI, and ICDAR KWS 2015. It is noticed that the proposed method outperforms state-of-the-art learning-free KWS methods considered here for comparison while evaluated on the present datasets. We also evaluate the performance of the present KWS model using state-of-the-art deep features and it is found that the features used in the present work perform better than the deep features extracted using InceptionV3, VGG19, and DenseNet121 models. Full article

One of the biometric methods in authentication systems is the writer verification/identification using password handwriting. The main objective of this paper is to present a robust writer verification system by using cursive texts as well as block letter words. To evaluate the system, two datasets have been used. One of them is called Secure Password DB 150, which is composed of 150 users with 18 samples of single character words per user. Another dataset is public and called IAM online handwriting database, and it is composed of 220 users of cursive text samples. Each sample has been defined by a set of features, composed of 67 geometrical, statistical, and temporal features. In order to get more discriminative information, two feature reduction methods have been applied, Fisher Score and Info Gain Attribute Evaluation. Finally, the classification system has been implemented by hold-out cross validation and k-folds cross validation strategies for three different classifiers, K-NN, Naïve Bayes and Bayes Net classifiers. Besides, it has been applied for verification and identification approaches. The best results of 95.38% correct classification are achieved by using the k-nearest neighbor classifier for single character DB. A feature reduction by Info Gain Attribute Evaluation improves the results for Naïve Bayes Classifier to 98.34% for IAM online handwriting DB. It is concluded that the set of features and its reduction are a strong selection for the based-password handwritten writer identification in comparison with the state-of-the-art. Full article