Level Features Research Articles

GFS-YOLO11: A Maturity Detection Model for Multi-Variety Tomato

In order to solve the problems that existing tomato maturity detection methods struggle to take into account both common tomato and cherry tomato varieties in complex field environments (such as light change, occlusion, and fruit overlap) and the model size being too large, this paper proposes a lightweight tomato maturity detection model based on improved YOLO11, named GFS-YOLO11. In order to achieve a lightweight network, we propose the C3k2_Ghost module to replace the C3K2 module in the original network, which can ensure a feature extraction capability and reduce model computation. In order to compensate for the potential feature loss caused by the light weight, this paper proposes a feature-refining module (FRM). After embedding each feature extraction module in the trunk network, it improves the feature expression ability of common tomato and cherry tomato in complex field environments by means of depth-separable convolution, multi-scale pooling, and channel attention and spatial attention mechanisms. In addition, in order to further improve the detection ability of the model for tomatoes of different sizes, the SPPFELAN module is also proposed in this paper. In combining the advantages of SPPF and ELAN, multiple parallel SPPF branches are used to extract features of different levels and perform splicing and fusion. To verify the validity of the method, this study constructed a dataset of 1061 images of common and cherry tomatoes, covering tomatoes in six ripened categories. The experimental results show that the performance of the GFS-YOLO11 model is significantly improved compared with the original model; the P, R, mAP50, and MAP50-95 increased by 5.8%, 4.9%, 6.2%, and 5.5%, respectively, and the number of parameters and calculation amount were reduced by 35.9% and 22.5%, respectively. The GFS-YOLO11 model is lightweight while maintaining high precision, can effectively cope with complex field environments, and more conveniently meet the needs of real-time maturity detection of common tomatoes and cherry tomatoes.

Role of Ultrasonographic and Clinicopathological Characteristics in Assessing Stromal Tumor-Infiltrating Lymphocyte Density in Triple-Negative Breast Cancer

A high density of stromal tumor-infiltrating lymphocytes (sTILs) is positively correlated with the pathological complete response rate and favorable survival in patients with triple-negative breast cancer (TNBC). Heterogeneity in stromal lymphocyte distribution and limited tumor sampling may affect the accuracy of sTIL quantification in biopsy samples from patients receiving neoadjuvant therapy. Thus, identifying additional biomarkers to complement sTIL evaluation is essential. To identify biomarkers that could be used to complement sTILs evaluation. Retrospective cohort study. A total of 162 patients with invasive TNBC were enrolled in the study. The following data were gathered: sTIL density, Ki67, nuclear grades of cancer cells, lymphovascular invasion status, the American Joint Committee on Cancer stage, axillary lymph node metastasis status, and ultrasonographic parameters of the tumor (size, shape, orientation, margin, internal echo pattern, posterior feature, and vascularity). The relationship between sTIL density and ultrasonographic or clinicopathological characteristics was investigated using both continuous and categorical analyses. Posterior features of the primary tumors was associated with sTIL density (p = 0.038). Additionally, the Ki67 levels and nuclear grades were also associated with sTIL density (p < 0.001 and p = 0.024, respectively). When stratified according to a 20% cut-off of sTIL density, the posterior features of primary tumors, Ki67 levels, and nuclear grades significantly differed between the high and low sTIL density tumors. Tumors with high Ki67 levels were more likely to exhibit high sTIL density than low sTIL density [odds ratio (OR): 2.75, p = 0.021]. Furthermore, nuclear grade III tumors demonstrated significantly higher sTIL density than nuclear grade I-II tumors (OR: 2.49, p = 0.014). Additionally, tumors with posterior enhancement or no posterior features were more likely to exhibit high sTIL density than tumors with acoustic shadows (OR: 2.91, p = 0.028; OR: 2.74, p = 0.022, respectively). Low sTIL density is frequently observed in tumors exhibiting acoustic shadows on ultrasound. However, high sTIL density is more common in tumors with posterior enhancement or no posterior features. Furthermore, high Ki67 levels (> 40%) and high nuclear grades are positively correlated with high sTIL density. This study findings highlight the need for closer surveillance of these biomarkers to complement sTIL evaluation in TNBC.

Glaucoma detection using hybrid machine learning techniques

An eye disease called glaucoma can cause irreversible blindness if left undetected and not properly managed. The major challenge is that glaucoma often has no symptoms in its early stages, making it hard to detect using traditional testing methods like eye pressure measurements and eye exams. Several techniques for Glaucoma detection encountered difficulties due to a short training dataset, resulting in overfitting and under fitting problems. A hybrid machine learning approach based on CNN-SVM for detection of Glaucoma was proposed. Initially images taken from glaucoma dataset were preprocessed using standard scalar, then the preprocessed images are fed into CNN for transformation into high level features, the extracted features are subsequently passed onto an SVM classifier to distinguish between normal and glaucomatous conditions. Experimental results for the proposed CNN-SVM offers an accuracy, precision, recall and F1-score of 100% demonstrating its superiority over other existing techniques such as SVM which has accuracy, precision, recall and F1-score of 93%, 92%, 90% and 94% respectively and CNN with the accuracy, precision, recall and F1- score of 95%, 99%, 88% and 90% respectively. The integration of CNN and SVM presents a promising framework for automated Glaucoma detection, offering significant potential for real-world clinical applications.

Recurrent Neural Networks based Online Behavioural Malware Detection Techniques for Cloud Infrastructure

: With the increased basis of applications and services on Cloud, the dependence on cloud technologies has expanded rapidly. This transition eases some of the hardware management troubles, but it also exposes broad security threats, especially in regard to malware. Malware is a major threat to cloud computing systems mainly in IaaS dominated environments. In this paper we explore the usage of Recurrent Neural Networks, RNNs in malware detection in cloud VMs. In particular, we examine two of the most popular RNN models; Long short term Memory, LSTM and Bidirectional RNNs, BIDI, that try to understand malware patterns by watching the activity of systems and the use of CPU, Memory and Disk. On the dataset of 40,680 samples 40,680 malware and benign samples, the models were trained on process level features from real malware in a free and open cloud environment. Both models achieved detection rates of over 99% in average key performance metrics. Moreover, wealso investigate the impact of different representations of input data on the performance of the models. The results of the study clearly support the viability of RNNs for online malware detection in cloud infrastructure, and allow for high accuracy and real time detection.

LINGUOCOGNITIVE AND LINGUOCULTURAL FOUNDATIONS OF THE PROCESSES OF TERMINOLOGIZATION IN MODERN KAZAKH AND TURKISH LANGUAGES

The intensification of the activity of the Kazakh language in accordance with its state status in all spheres of society, the renewal of its potential strengthened the connection of the language with the form of linguistic consumption, began to influence the requirements of everyday practice. This is evident from the active process of terminology creation in the social, economic, political, scientific and technical spheres. In addition, the core of the neonominational process, which created the interrelationships of language~ society, language~consciousness, language ~ culture, etc., is not only scientific and informational, but also cultural and social content. This is the basis for improving not only the quantitative, but also the qualitative level of modern language development. Therefore, the recognition of the influence of language and culture on each other in the form of new terms and the function of language as a unique means of transmitting culture within the framework of cultural representation is especially relevant. Based on this, new names and terms in the Kazakh language, which also serve as a unique feature of the cognitive level of language users, are considered in the article as a “linguistic picture of the world” in accordance with the process of updating the language. In this regard, the purpose of this article is to compare and characterize the linguocognitive and linguocultural foundations of terminologization processes in modern Kazakh and Turkish languages. The following methods were used in the course of the study: descriptive, comparative, cognitive, ethnolinguistic and conceptual. Also, to determine the cultural and national component structured in the mind of a native speaker in a terminologized word, diachronic, contrasting, definitive were used, and discursive analysis, connotative approach, and motivation methods were used to recognize the process of terminologization. The value of the research: by describing and showing the linguocognitive and linguocultural foundations of terminologization processes in modern Kazakh and Turkish languages, we contribute to the improvement and strengthening of the “Turkic world”. The linguistic data and conclusions presented in the article can be used as practical and theoretical material for research by students, undergraduates, doctoral students and young people studying at the faculties of Philology, Translation Studies, Oriental Studies and Turkology.

Pottery evolution pattern discovery based on deep learning: case study of Miaozigou culture in China

Potteries, one of the tools widely used by early humans, encapsulates rich historical information. Deep neural networks have been applied to analyzing pottery digital images, bypassing the need for intricate handcrafted features. However, existing models focus solely on pottery shape comparison, neglecting the analysis of their evolution across different historical periods. In this work, we propose a method based on deep learning to assist experts in identifying the evolutionary patterns of a given pottery type within their specified chronological divisions. First we train a convolutional neural network for pottery classification, extracting low and high level features that represent different ages of pottery samples. Next, we employ clustering algorithms to identify representative potteries for each historical period based on high level features. To facilitate intuitive comparisons across different ages, we use shallow features and compute cosine similarities between potteries, visualizing shape and decoration differences. This approach enhances understanding of pottery evolution patterns directly through visual analysis. The effectiveness and efficiency of our proposed method are evaluated by validating it on three distinct era division cases using data from the Dabagou and Miaozigou archaeological sites, which represent the Miaozigou culture and exhibit clear evolutionary patterns. Our method identifies representative artifacts for each era and uncovers their evolutionary patterns effectively and efficiently, achieving conclusions comparable to those of experts while significantly reducing time compared to traditional manual methods.

Understanding responses of people with ASD in diverse reasoning tasks: A formal study.

Recent studies have shown that in some reasoning tasks people with Autism Spectrum Disorder perform better than typically developing people. This paper compares four such tasks, namely a syllogistic task, two decision-making tasks, and a task from the heuristics and biases literature, the aim being to identify common structure as well as differences. In the terminology of David Marr's three levels of cognitive systems, the tasks show commonalities on the computational level in terms of the effect of contextual stimuli, though an in-depth analysis of such contexts provides certain distinguishing features in the algorithmic level. We also make some general remarks on our approach, so as to set the stage for further studies in the area which could provide a better understanding of the reasoning process of ASD individuals.

Specifying and Verifying Information Flow Control in SELinux Configurations

Security Enhanced Linux (SELinux) is a security architecture for Linux implementing Mandatory Access Control. It has been used in numerous security-critical contexts ranging from servers to mobile devices. However, its application is challenging as SELinux security policies are difficult to write, understand, and maintain. Recently, the intermediate language CIL was introduced to foster the development of high-level policy languages and to write structured configurations. Despite CIL’s high level features, CIL configurations are hard to understand as different constructs interact in non-trivial ways. Moreover, there is no mechanism to ensure that a given configuration obeys desired information flow policies. To remedy this, we enrich CIL with a formal semantics, and we propose IFCIL, a backward compatible extension of CIL for specifying fine-grained information flow requirements. Using IFCIL, administrators can express confidentiality, integrity, and non-interference properties. We also provide a tool to statically verify these requirements and we experimentally assess it on ten real-world policies.

A Comparative Analysis of Attention Mechanism in RNN-LSTMs for Improved Image Captioning Performance

Image captioning which links computer vision with NATURAL LANGUAGE PROCESSING is critical in providing descriptions for the image. The proposed solution in this research is a hierarchical attention model which includes use of CNN features on images and LSTM networks with attention mechanisms for generating captions. By utilizing both object level and image level features, our method enhances the quality and relevance of captions, enhancing the variability of the automated image description.

Grasping the Concept of an Object at a Glance: Category Information Accessed by Brief Dichoptic Presentation.

What type of conceptual information about an object do we get at a brief glance? In two experiments, we investigated the nature of conceptual tokening-the moment at which conceptual information about an object is accessed. Using a masked picture-word congruency task with dichoptic presentations at "brief" (50-60 ms) and "long" (190-200 ms) durations, participants judged the relation between a picture (e.g., a banana) and a word representing one of four property types about the object: superordinate (fruit), basic level (banana), a high-salient (yellow), or low-salient feature (peel). In Experiment 1, stimuli were presented in black-and-white; in Experiment 2, they were presented in red and blue, with participants wearing red-blue anaglyph glasses. This manipulation allowed for the independent projection of stimuli to the left- and right-hemisphere visual areas, aiming to probe the early effects of these projections in conceptual tokening. Results showed that superordinate and basic-level properties elicited faster and more accurate responses than high- and low-salient features at both presentation times. This advantage persisted even when the objects were divided into categories (e.g., animals, vegetables, vehicles, tools), and when objects contained high-salient visual features. However, contrasts between categories show that animals, fruits, and vegetables tend to be categorized at the superordinate level, while vehicles tend to be categorized at the basic level. Also, for a restricted class of objects, high-salient features representing diagnostic color information (yellow for the picture of a banana) facilitated congruency judgments to the same extent as that of superordinate and basic-level labels. We suggest that early access to object concepts yields superordinate and basic-level information, with features only yielding effects at a later stage of processing, unless they represent diagnostic color information. We discuss these results advancing a unified theory of conceptual representation, integrating key postulates of atomism and feature-based theories.

"Visual verbs": Dynamic event types are extracted spontaneously during visual perception.

During visual processing, input that is continuous in space and time is segmented, resulting in the representation of discrete tokens-objects or events. And there has been a great deal of research about how object representations are generalized into types-as when we see an object as an instance of a broader category (e.g., an animal or plant). There has been much less attention, however, to the possibility that vision represents dynamic information in terms of a small number of primitive event types (such as twisting or bouncing). (In models that posit a "language of vision," these would be the foundational visual verbs.) Here we ask whether such event types are extracted spontaneously during visual perception, even when entirely task irrelevant during passive viewing. We exploited the phenomenon of categorical perception-wherein differences are more readily noticed when they are represented in terms of different underlying categories. Observers were better at detecting changes to images or short videos when the changes involved switches in the underlying event type-even when the changes that maintained the same event type were objectively larger (in terms of both brute image metrics and higher level feature change). We observed this categorical "cross-event-type" advantage for visual working memory for twisting versus rotating, scooping versus pouring, and rolling versus bouncing. Moreover, additional control experiments confirmed that such effects could not be explained by appeal to lower-level non-categorical stimulus differences. This spontaneous perception of "visual verbs" might promote both generalization and prediction about how events are likely to unfold. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

R-Net: Recursive decoder with edge refinement network for salient object detection

Fully convolutional neural (FCN) networks based on encoder-decoder structures are adopted for salient object detection (SOD) and achieve state-of-the-art performance. However, most of the previous works aggregate multilevel features layer by layer and ignore long-range dependencies of spatial context, leading to imbalanced feature processes and context separation. In addition, the blurry edges obtained from previous SOD methods inevitably influence the detection accuracy. To handle these problems, this study proposes an effective recursion-based SOD network R-Net adopting a partial cascade structure, which contains a recursive decoder module (RDM), a long-range dependency module (LRDM), and an edge refinement module (ERM). The RDM is composed of three subdecoders, which can scale different level features to the same size through recursive pooling and recursive upsampling for full-scale feature fusion. The LRDM bridges the encoder and decoder by weighting multilevel features to establish the long-range feature dependency of the spatial context. And the ERM attached to the decoder introduces the shallow reference feature to refine the blurry edges for obtaining delicate detection results. The SOD experimental results on DUTS-TE, HKU-IS, PASCAL-S, ECSSD, and DUT-OMRON datasets demonstrate that the proposed R-Net is more robust under different complex scenes compared with some state-of-the-art methods. The source codes are available at https://github.com/ZEROICEWANG/R-Net.

Scene-aware classifier and re-detector for thermal infrared tracking

Compared with common visible light scenes, the target of infrared scenes lacks information such as the color, texture. Infrared images have low contrast, which not only lead to interference between targets, but also interference between the target and the background. In addition, most infrared tracking algorithms lack a redetection mechanism after lost target, resulting in poor tracking effect after occlusion or blurring. To solve these problems, we propose a scene-aware classifier to dynamically adjust low, middle, and high level features, improving the ability to utilize features in different infrared scenes. Besides, we designed an infrared target re-detector based on multi-domain convolutional network to learn from the tracked target samples and background samples, improving the ability to identify the differences between the target and the background. The experimental results on VOT-TIR2015, VOT-TIR2017 and LSOTB-TIR show that the proposed algorithm achieves the most advanced results in the three infrared object tracking benchmark.

Zero shot plant disease classification with semantic attributes

In the rapidly evolving field of plant disease detection, the number and complexity of crop diseases are increasing, made worse by factors like climate change. Addressing these challenges requires robust and efficient methodologies capable of early and accurate disease identification. This paper explores the integration of advanced deep learning techniques, including pre-trained models, zero-shot learning, and semantic attributes to enhance the effectiveness of plant disease detection systems. High level features extracted from the images by these pretrained models capture crucial patterns, while domain-specific semantic attributes, such as leaf texture and color variations, enhance the understanding. Incorporating zero-shot learning enables adaptation to new and unseen diseases using semantic descriptions. Experimental validation across diverse plant species and disease types underscores the approach’s reliability in real-world agricultural scenarios. Our approach has demonstrated superior performance with plant village dataset, showing a significant improvement in accuracy and generalization. These results underscore the potential of our method to revolutionize plant disease detection and management in agricultural practices.

TAFENet: A Two-Stage Attention-Based Feature-Enhancement Network for Strip Steel Surface Defect Detection

Strip steel serves as a crucial raw material in numerous industries, including aircraft and automobile manufacturing. Surface defects in strip steel can degrade the performance, quality, and appearance of industrial steel products. Detecting surface defects in steel strip products is challenging due to the low contrast between defects and background, small defect targets, as well as significant variations in defect sizes. To address these challenges, a two-stage attention-based feature-enhancement network (TAFENet) is proposed, wherein the first-stage feature-enhancement procedure utilizes an attentional convolutional fusion module with convolution to combine all four-level features and then strengthens the features of different levels via a residual spatial-channel attention connection module (RSC). The second-stage feature-enhancement procedure combines three-level features using an attentional self-attention fusion module and then strengthens the features using a RSC attention module. Experiments on the NEU-DET and GC10-DET datasets demonstrated that the proposed method significantly improved detection accuracy, thereby confirming the effectiveness and generalization capability of the proposed method.

A multi-task deep learning model based on comprehensive feature integration and self-attention mechanism for predicting response to anti-PD1/PD-L1

BackgroundImmune checkpoint inhibitor (ICI) has been widely used in the treatment of advanced cancers, but predicting their efficacy remains challenging. Traditional biomarkers are numerous but exhibit heterogeneity within populations. For comprehensively utilizing the ICI-related biomarkers, we aim to conduct multidimensional feature selection and deep learning model construction. MethodsWe used statistical and machine learning methods to map features of different levels to next-generation sequencing gene expression. We integrated genes from different sources into the feature input of a deep learning model, by means of self-attention mechanism. ResultsWe performed feature selection at the single-cell sequencing level, PD-L1 (CD274) analysis level, tumor mutational burden (TMB)/mismatch repair (MMR) level, and somatic copy number alteration (SCNA) level, obtaining 96 feature genes. Based on the pan-cancer dataset, we trained a multi-task deep learning model. We tested the model in the bladder urothelial carcinoma testing set 1 (AUC = 0.62, n = 298), bladder urothelial carcinoma testing set 2 (AUC = 0.66, n = 89), non-small cell lung cancer testing set (AUC = 0.85, n = 27), and skin cutaneous melanoma testing set (AUC = 0.71, n = 27). ConclusionOur study demonstrates the potential of the deep learning model for integrating multidimensional features in predicting the outcome of ICI. Our study also provides a potential methodological case for medical scenarios requiring the integration of multiple levels of features.

A HybridOpt approach for early Alzheimer’s Disease diagnostics with Ant Lion Optimizer (ALO)

Alzheimer disease is a neurological disorder that affects the elderly, caused by abnormal protein buildup in the brain. It leads to difficulties such as financial mismanagement, disorientation, behavioral changes, and repetitive speech. The existing methods use traditional features to detect the early signs of AD with low detection accuracy. The potential features have to be identified that represent best the patterns associated with alzheimers. Feature selection using ant lion optimization resolves the issue by using complementarty information from hybrid features. The proposed HybridOpt pipeling for AD diagnosis combines the high level and low level features for early stage detection The objective of this work is to select efficient features from different deep networks, such as AlexNet, Googlenet, VGG16, ResNet, Efficient, DenseNet, and traditional texture features. Ant Lion Optimization is used to select the best feature among the deep network and traditional texture feature groups. Extensive experimentation on two highly challenging datasets called the Alzheimer’s disease neuroimage dataset and KAGGLE reveals that the proposed HybridOpt pipeline achieves an accuracy of 99% and 98.1% respectively.

PCP-GC-LM: single-sequence-based protein contact prediction using dual graph convolutional neural network and convolutional neural network

BackgroundRecently, the process of evolution information and the deep learning network has promoted the improvement of protein contact prediction methods. Nevertheless, still remain some bottleneck: (1) One of the bottlenecks is the prediction of orphans and other fewer evolution information proteins. (2) The other bottleneck is the method of predicting single-sequence-based proteins mainly focuses on selecting protein sequence features and tuning the neural network architecture, However, while the deeper neural networks improve prediction accuracy, there is still the problem of increasing the computational burden. Compared with other neural networks in the field of protein prediction, the graph neural network has the following advantages: due to the advantage of revealing the topology structure via graph neural network and being able to take advantage of the hierarchical structure and local connectivity of graph neural networks has certain advantages in capturing the features of different levels of abstraction in protein molecules. When using protein sequence and structure information for joint training, the dependencies between the two kinds of information can be better captured. And it can process protein molecular structures of different lengths and shapes, while traditional neural networks need to convert proteins into fixed-size vectors or matrices for processing.ResultsHere, we propose a single-sequence-based protein contact map predictor PCP-GC-LM, with dual-level graph neural networks and convolution networks. Our method performs better with other single-sequence-based predictors in different independent tests. In addition, to verify the validity of our method against complex protein structures, we will also compare it with other methods in two homodimers protein test sets (DeepHomo test dataset and CASP-CAPRI target dataset). Furthermore, we also perform ablation experiments to demonstrate the necessity of a dual graph network. In all, our framework presents new modules to accurately predict inter-chain contact maps in protein and it’s also useful to analyze interactions in other types of protein complexes.

MGFmiRNAloc: Predicting miRNA Subcellular Localization Using Molecular Graph Feature and Convolutional Block Attention Module.

MiRNA has distinct physiological functions at various cellular locations. However, few effective computational methods for predicting the subcellular location of miRNA exist, thereby leaving considerable room for improvement. Accordingly, our study proposes the MGFmiRNAloc simplified molecular input line entry system (SMILES) format as a new approach for predicting the subcellular localization of miRNA. Additionally, the graphical convolutional network (GCN) technique was employed to extract the atomic nodes and topological structure of a single base, thereby constructing RNA sequence molecular map features. Subsequently, the channel attention and spatial attention mechanisms (CBAM) were designed to mine deeper for more efficient information. Finally, the prediction module was used to detect the subcellular localization of miRNA. The 10-fold cross-validation and independent test set experiments demonstrate that MGFmiRNAloc outperforms the most sophisticated methods. The results indicate that the new atomic level feature representation proposed in this study could overcome the limitations of small samples and short miRNA sequences, accurately predict the subcellular localization of miRNAs, and be extended to the subcellular localization of other sequences.

Analysis of unsteady wake flow in centrifugal pump volute by using mode decomposition method

To analyze the coherent structure of wake flow in volute and its corresponding frequency information, dynamic modal decomposition (DMD), classic and spectral proper orthogonal decomposition (SPOD), were employed to decompose the transient flow field of the centrifugal pump volute. The snapshot set was constructed by means of the velocity field data at different moments in volute based on the large eddy simulation (LES) approach. The basic principles of the DMD, POD, and SPOD methods were compared in detail, and the decomposition results of the three methods on the wake flow structure in volute were compared and analyzed. The analysis results show that DMD can decompose the wake flow into coherent structures with different frequencies, including the basic steady-state structure, the dynamic modal flow field structure characterizing rotor–stator interaction (the first three modes with frequencies of 145.81, 291.61, and 437.43 Hz, respectively), and dissipative modal flow field structure characterizing fragmentized vortex (the fourth mode with a frequency of 486.03 Hz) in the volute. The POD can decompose the wake flow into flow structures with different energy levels. The first four modal energies account for more than 66% of total energy, which represents the large-scale structure with higher energy, and its dominant frequencies correspond to the blade passing frequency (145 Hz) and its frequency multiplication (290 Hz). The SPOD can not only decompose the complex wake flow into structural features of different energy levels but also has single-frequency characteristics of its modal structures. Compared with DMD and POD methods, the SPOD has the advantages of both, and can reflect the evolution characteristics of the wake flow in the volute.