Improve Detection Performance Research Articles

Event Detection and Classification in Tweets using Deep Learning

Online social networks have become important sources of information and contextual data in all areas of life, including finance, elections, social events, health, sports, etc. Recently, the detection and classification of useful events presented in tweets has attracted a lot of interest. However, due to the inherent challenges associated with the nature of the events to be detected or classified, traditional approaches have not yielded satisfactory results. The use of deep learning-based text word embedding representations, such as Word2Vec, GloVe, FastText, and BERT, has shown significant efficacy in improving detection performance by considering the semantic context. This study proposes a model that uses an LSTM stacked on top of BERT representations to effectively detect and classify events in tweets. To this end, a dataset of about 310,000 event-related tweets has been collected and categorized into 50 event types based on a selected set of representative keywords. Multiple experiments were carried out on the collected dataset to evaluate the performance of the proposed model. The proposed model attained an overall accuracy greater than 94.3% and an F1 score of more than 90%, achieving state-of-the-art results in the classification of most of the event categories.

Visible light-responsive enrofloxacin PEC aptasensor based on CN QDs sensitized Bi4O5Br2 nanosheets.

The excessive application of enrofloxacin (ENR) results in residues contaminating both food and the environment. Consequently, developing robust analytical methods for the selective detection of ENR is crucial. The photoelectrochemical (PEC) sensor has emerged as a highly sensitive analytical technique that has seen rapid development in recent years. The functioning of a PEC sensor relies on the reducing capacity of photogenerated electrons and the oxidizing capacity of photogenerated holes produced by the photoactive material. Bi4O5Br2 demonstrates its potential in electrochemical detection, but faces inherent challenges, including swift electron-hole recombination and slow carrier migration, which hinder its catalytic activity. In this study, we synthesized carbon nitride quantum dots doped with Bi4O5Br2 (CN QDs/Bi4O5Br2) through an in situ growth method, utilizing this composite as a photoactive material. The incorporation of CN QDs leads to a 17-fold increase in photocurrent compared to Bi4O5Br2 alone. This enhancement is attributed not only to the improved separation of electron-hole pairs, facilitated by the CN QDs, which boosts photocatalytic activity, but also to the enlarged range of visible light absorption. We employed an ENR-specific aptamer as the recognition element, resulting in the construction of a high-performance photoelectrochemical aptasensor for ENR detection. The sensor exhibited a linear detection range of 1×10-1 to 1×106ngmL-1 and a detection limit of 0.033ngmL-1. The impressive performance of the CN QDs/Bi4O5Br2 sensing platform demonstrates its potential application in detecting ENR concentrations in food, biomedical contexts, and environmental analyses. Benefiting from the sensitization of CN QDs, CN QDs/Bi4O5Br2 exhibited 17-fold PEC signal of pure Bi4O5Br2. The presence of quantum dots in CN QDs/Bi4O5Br2 facilitates rapid separation of electron-hole pairs, leading to significantly enhanced PEC activity and improved detection performance for ENR. This research convincingly illustrates that integrating CN QDs with Bi4O5Br2 nanosheets could pave the way for designing more efficient bismuth-based semiconductor photoactive materials for sensing applications.

Multi-Scale Hierarchical Feature Fusion for Infrared Small-Target Detection

Detecting small targets in infrared images presents significant challenges due to their tiny size and complex backgrounds, making this task a hotspot for research. Traditional methods rely on assumption-based modeling and manual design, struggling to handle the variability of real-world scenarios. Although convolutional neural networks (CNNs) increase robustness to diverse scenes with a data-driven paradigm, many CNN-based methods are insufficient in capturing fine-grained details necessary for small targets and are less effective during multi-scale feature fusion. To overcome these challenges, we propose the novel Wide-scale Gated Fully Fusion Network (WGFFNet) in this article, which contributes to infrared small-target detection (IRSTD). WGFFNet uses a classic encoder–decoder structure, where the designed stepped fusion block (SFB) embedded in the feature extraction stage captures finer local context across multiple scales during encoding, and along the decoding path, the multi-level features are progressively integrated by a Fully Gated Interaction (FGI) Module to enhance feature representation. The inclusion of a boundary difference loss further optimizes the edge details of targets. We conducted comprehensive experiments on two public infrared small-target datasets: SIRST-V2 and IRSTD-1k. Quantitative and qualitative results demonstrate that our WGFFNet outperforms representative methods when considering various evaluation metrics together, achieving an improved detection performance and computational efficiency for detecting small targets in infrared images.

Transformer Architecture for Micromotion Target Detection Based on Multi-Scale Subaperture Coherent Integration

In recent years, long-time coherent integration techniques have gained significant attention in maneuvering target detection due to their ability to effectively enhance the signal-to-noise ratio (SNR) and improve detection performance. However, for space targets, challenges such as micromotion phenomena and complex scattering characteristics make envelope alignment and phase compensation difficult, thereby limiting integration gain. To address these issues, in this study, we conducted an in-depth analysis of the echo model of cylindrical space targets (CSTs) based on different types of scattering centers. Building on this foundation, the multi-scale subaperture coherent integration Transformer (MsSCIFormer) was proposed, which integrates MsSCI with a Transformer architecture to achieve precise detection and motion parameter estimation of space targets in low-SNR environments. The core of the method lies in the introduction of a convolutional neural network (CNN) feature extractor and a dual-attention mechanism, covering both intra-subaperture attention (Intra-SA) and inter-subaperture attention (Inter-SA). This design efficiently captures the spatial distribution and motion patterns of the scattering centers of space targets. By aggregating multi-scale features, MsSCIFormer significantly enhances the detection performance and improves the accuracy of motion parameter estimation. Simulation experiments demonstrated that MsSCIFormer outperforms traditional moving target detection (MTD) methods and other deep learning-based algorithms in both detection and estimation tasks. Furthermore, each module proposed in this study was proven to contribute positively to the overall performance of the network.

DCFA-YOLO: A Dual-Channel Cross-Feature-Fusion Attention YOLO Network for Cherry Tomato Bunch Detection

To better utilize multimodal information for agriculture applications, this paper proposes a cherry tomato bunch detection network using dual-channel cross-feature fusion. It aims to improve detection performance by employing the complementary information of color and depth images. Using the existing YOLOv8_n as the baseline framework, it incorporates a dual-channel cross-fusion attention mechanism for multimodal feature extraction and fusion. In the backbone network, a ShuffleNetV2 unit is adopted to optimize the efficiency of initial feature extraction. During the feature fusion stage, two modules are introduced by using re-parameterization, dynamic weighting, and efficient concatenation to strengthen the representation of multimodal information. Meanwhile, the CBAM mechanism is integrated at different feature extraction stages, combined with the improved SPPF_CBAM module, to effectively enhance the focus and representation of critical features. Experimental results using a dataset obtained from a commercial greenhouse demonstrate that DCFA-YOLO excels in cherry tomato bunch detection, achieving an mAP50 of 96.5%, a significant improvement over the baseline model, while drastically reducing computational complexity. Furthermore, comparisons with other state-of-the-art YOLO and other object detection models validate its detection performance. This provides an efficient solution for multimodal fusion for real-time fruit detection in the context of robotic harvesting, running at 52fps on a regular computer.

2-Propanol Suspension Method to Increase Acetylcholinesterase and Flow Stability on μPADs.

Ensuring detection performance and shelf life is crucial for analytical devices. Advances in materials and reaction mechanisms have improved detection performance, yet extending the operational lifetime of microfluidic paper-based analytical devices (μPADs)─especially those reliant on sensitive enzymes─remains a challenge. Here, we present an alternative to air-drying and lyophilization: loading enzymes suspended in 2-propanol (iPrOH). By suspending the enzyme in iPrOH, we circumvent the enzyme activity losses commonly associated with freeze-thawing and freeze-drying. Accelerated aging tests, supported by statistical analyses of long-term activity retention (including comparisons over multiple time points), indicate that while conventional methods do not sustain consistent superiority, the iPrOH suspension method maintains higher enzymatic activity over extended periods. By avoiding stabilizers and circumventing the limitations of other techniques, our method enables μPADs to achieve both longevity and stable fluid flow. Thus, we provide a more robust, on-site analytical platform capable of reliable on-site detection.

Integrating EnlightenGAN for enhancing car logo detection under challenging lighting conditions

Abstract Detecting small objects in challenging lighting conditions is a critical task with broad applications in computer vision, surveillance, and medical imaging. Small objects are especially vulnerable to detection errors due to their size and the significant impact of varying lighting conditions, such as low-light environments or high glare. Furthermore, the lack of a benchmark dataset that accurately simulates real-world scenarios makes it difficult to evaluate the performance of detection models in such conditions. In the proposed paper, we address these challenges by developing a comprehensive dataset specifically designed for car logo detection under diverse and challenging lighting environments. The dataset incorporates variations in lighting, viewing angles, and object scales to replicate real-world conditions. We also introduce a detection pipeline integrating EnlightenGAN for low-light image enhancement, which preserves fine details and textures, and YOLOv7 and YOLOv8 for improved detection accuracy. By enhancing image quality and refining object detection models, our approach significantly boosts the detection of small car logos, even in challenging lighting scenarios where traditional models tend to fail. Experimental results show that YOLOv8 outperforms YOLOv7 by approximately 7%, with YOLOv8 achieving a mean Average Precision (mAP) of 73.3% on the augmented dataset compared to 68.4% for the original dataset and 67.2% for the processed images. These results validate the effectiveness of GAN-based enhancement techniques in preserving small object details and improving detection performance. Additionally, introducing a new dataset and benchmark provides a valuable resource for future research in small object detection under complex lighting conditions.

DCTCNet: Sequency discrete cosine transform convolution network for visual recognition.

The discrete cosine transform (DCT) has been widely used in computer vision tasks due to its ability of high compression ratio and high-quality visual presentation. However, conventional DCT is usually affected by the size of transform region and results in blocking effect. Therefore, eliminating the blocking effects to efficiently serve for vision tasks is significant and challenging. In this paper, we introduce All Phase Sequency DCT (APSeDCT) into convolutional networks to extract multi-frequency information of deep features. Due to the fact that APSeDCT can be equivalent to convolutional operation, we construct corresponding convolution module called APSeDCT Convolution (APSeDCTConv) that has great transferability similar to vanilla convolution. Then we propose an augmented convolutional operator called MultiConv with APSeDCTConv. By replacing the last three bottleneck blocks of ResNet with MultiConv, our approach not only reduces the computational costs and the number of parameters, but also exhibits great performance in classification, object detection and instance segmentation tasks. Extensive experiments show that APSeDCTConv augmentation leads to consistent performance improvements in image classification on ImageNet across various different models and scales, including ResNet, Res2Net and ResNext, and achieving 0.5%-1.1% and 0.4%-0.7% AP performance improvements for object detection and instance segmentation, respectively, on the COCO benchmark compared to the baseline.

Improved YOLOv10 for Visually Impaired: Balancing Model Accuracy and Efficiency in the Case of Public Transportation

Advancements in automation and artificial intelligence have significantly impacted accessibility for individuals with visual impairments, particularly in the realm of bus public transportation. Effective bus detection and bus point-of-view (POV) classification are crucial for enhancing the independence of visually impaired individuals. This study introduces the Improved-YOLOv10, a novel model designed to tackle challenges in bus identification and pov classification by integrating Coordinate Attention (CA) and Adaptive Kernel Convolution (AKConv) into the YOLOv10 framework. The Improved YOLOv10 advances the YOLOv10 architecture through the incorporation of CA, which enhances long-range dependency modeling and spatial awareness, and AKConv, which dynamically adjusts convolutional kernels for superior feature extraction. These enhancements aim to improve both detection accuracy and efficiency, essential for real-time applications in assistive technologies. Evaluation results demonstrate that the Improved-YOLOv10 offers significant improvements in detection performance, including better Accuracy, Precision and Recall compared to YOLOv10. The model also exhibits reduced computational complexity and storage requirements, highlighting its efficiency. While the classification results show some trade-offs, with slightly decreased overall F1 score, the complexity of Giga Floating Point Operations (GFLOPs), Parameters, and Weight/MB in the Improved-YOLOv10 remains advantageous for classification tasks. The model’s architectural improvements contribute to its robustness and efficiency, making it a suitable choice for real-time applications and assistive technologies.

DCFE-YOLO: A novel fabric defect detection method.

Accurate detection of fabric defects is crucial for quality control in the textile industry. However, the task of fabric defect detection remains highly challenging due to the complex textures and diverse defect patterns. To address the issues of inaccurate localization and false positives caused by complex textures and varying defect sizes, this paper proposes an improved YOLOv8-based fabric defect detection method. First, Dynamic Snake Convolution is introduced into the backbone network to enhance sensitivity to elongated and subtle defects, improving the extraction of edge and texture details. Second, a Channel Priority Convolutional Attention mechanism is incorporated after the Spatial Pyramid Pooling layer to enable more precise defect localization by leveraging multi-scale structures and channel priors. Finally, the feature fusion network integrates Partial Convolution and Efficient Multi-scale Attention, optimizing the fusion of information across different feature levels and spatial scales, which enhances the richness and accuracy of feature representations while reducing computational complexity. Experimental results demonstrate a significant improvement in detection performance. Specifically, mAP@0.5 increased by 2.9%, precision improved by 3.5%, and mAP@0.5:0.95 rose by 2.3%, highlighting the model's superior capability in detecting complex defects. The project is available at https://github.com/lilian998/fabric.

Comparison of Dimensionality Reduction Techniques to Improve Performance and Efficiency of Logistic Regression in Network Anomaly Detection

Network anomaly detection is a crucial process to identify abnormal network traffic, which may pose a security threat. This research aims to improve the performance and efficiency of Logistic Regression (LR) in network anomaly detection by applying dimension reduction techniques, such as Principal Component Analysis (PCA), Truncated Singular Value Decomposition (TSVD), t-Distributed Stochastic Neighbor Embedding (t-SNE), and Independent Component Analysis (ICA). The performance of each dimension reduction method is evaluated based on accuracy, precision, recall, F1-score, and computation time. The results show that TSVD provides the best performance with 95.86% accuracy, 0.96 precision, 0.96 recall, 0.95 F1-score, and 13.83 seconds computation time. In contrast, ICA showed the worst performance, especially in precision, recall, and F1-score, with values of 0.73, 0.83, and 0.78, respectively. Meanwhile, although t-SNE produces competitive accuracy, it has a high computational cost with an execution time of 1698.54 seconds. These findings show that choosing the right dimension reduction algorithm not only improves detection performance but also supports data processing efficiency, making it highly relevant for large-scale network security scenarios. Keywords: dimensionality reduction, Logistic Regression, network anamoly detection, performance evaluation, Truncated Singular Value Decomposition.

Enhanced Detection Performance of Acute Vertebral Compression Fractures Using a Hybrid Deep Learning and Traditional Quantitative Measurement Approach: Beyond the Limitations of Genant Classification.

This study evaluated the applicability of the classical method, height loss ratio (HLR), for identifying major acute compression fractures in clinical practice and compared its performance with deep learning (DL)-based VCF detection methods. Additionally, it examined whether combining the HLR with DL approaches could enhance performance, exploring the potential integration of classical and DL methodologies. End-to-End VCF Detection (EEVD), Two-Stage VCF Detection with Segmentation and Detection (TSVD_SD), and Two-Stage VCF Detection with Detection and Classification (TSVD_DC). The models were evaluated on a dataset of 589 patients, focusing on sensitivity, specificity, accuracy, and precision. TSVD_SD outperformed all other methods, achieving the highest sensitivity (84.46%) and accuracy (95.05%), making it particularly effective for identifying true positives. The complementary use of DL methods with HLR further improved detection performance. For instance, combining HLR-negative cases with TSVD_SD increased sensitivity to 87.84%, reducing missed fractures, while combining HLR-positive cases with EEVD achieved the highest specificity (99.77%), minimizing false positives. These findings demonstrated that DL-based approaches, particularly TSVD_SD, provided robust alternatives or complements to traditional methods, significantly enhancing diagnostic accuracy for acute VCFs in clinical practice.

XI2S-IDS: An Explainable Intelligent 2-Stage Intrusion Detection System

The rapid evolution of technologies such as the Internet of Things (IoT), 5G, and cloud computing has exponentially increased the complexity of cyber attacks. Modern Intrusion Detection Systems (IDSs) must be capable of identifying not only frequent, well-known attacks but also low-frequency, subtle intrusions that are often missed by traditional systems. The challenge is further compounded by the fact that most IDS rely on black-box machine learning (ML) and deep learning (DL) models, making it difficult for security teams to interpret their decisions. This lack of transparency is particularly problematic in environments where quick and informed responses are crucial. To address these challenges, we introduce the XI2S-IDS framework—an Explainable, Intelligent 2-Stage Intrusion Detection System. The XI2S-IDS framework uniquely combines a two-stage approach with SHAP-based explanations, offering improved detection and interpretability for low-frequency attacks. Binary classification is conducted in the first stage followed by multi-class classification in the second stage. By leveraging SHAP values, XI2S-IDS enhances transparency in decision-making, allowing security analysts to gain clear insights into feature importance and the model’s rationale. Experiments conducted on the UNSW-NB15 and CICIDS2017 datasets demonstrate significant improvements in detection performance, with a notable reduction in false negative rates for low-frequency attacks, while maintaining high precision, recall, and F1-scores.

Enhancing lesion detection in automated breast ultrasound using unsupervised multi-view contrastive learning with 3D DETR.

The inherent variability of lesions poses challenges in leveraging AI in 3D automated breast ultrasound (ABUS) for lesion detection. Traditional methods based on single scans have fallen short compared to comprehensive evaluations by experienced sonologists using multiple scans. To address this, our study introduces an innovative approach combining the multi-view co-attention mechanism (MCAM) with unsupervised contrastive learning. Rooted in the detection transformer (DETR) architecture, our model employs a one-to-many matching strategy, significantly boosting training efficiency and lesion recall metrics. The model integrates MCAM within the decoder, facilitating the interpretation of lesion data across diverse views. Simultaneously, unsupervised multi-view contrastive learning (UMCL) aligns features consistently across scans, improving detection performance. When tested on two multi-center datasets comprising 1509 patients, our approach outperforms existing state-of-the-art 3D detection models. Notably, our model achieves a 90.3% cancer detection rate with a false positive per image (FPPI) rate of 0.5 on the external validation dataset. This surpasses junior sonologists and matches the performance of seasoned experts.

Impurity detection of premium green tea based on improved lightweight deep learning model.

Tea may be mixed with impurities during picking and processing, which can lower their quality. At present, the sorting of impurities in premium green tea mainly relies on manual labor, which is inefficient. In response to the technical challenges in this industry, this article uses deep learning technology to detect impurities in premium green tea. However, in actual production, computing resources are limited by equipment, and large models are not suitable for deployment. On this basis, a lightweight model for detecting impurities in premium green tea is proposed. We conducted experiments on You Only Look Once version 8 (YOLOv8) models on a self-made dataset and selected the basic model and the teacher model. By replacing the loss function and lightweight convolution, model pruning, and knowledge distillation, the model achieves lightweighting while improving detection performance. The experimental results reveal that the Giga floating point operations (GFLOPs), parameters, precision (P), recall (R), mean average precision (mAP), and frames per second (FPS) of the improved model are 4.2, 791966 B, 0.9379, 0.8959, 0.9484, and 1362.7, respectively. Compared with the original model, the P, R, mAP, and FPS are improved by 0.0216, 0.0320, 0.0261, and 368.0, respectively. The GFLOPs and parameters are reduced by 3.9 (48.15%) and 2214462 B (73.66%), respectively. This study provides technical support for the intelligent sorting of impurities in premium green tea.

Underwater moving target detection using online robust principal component analysis and multimodal anomaly detection.

In shallow water, reverberation complicates the detection of low-intensity, variable-echo moving targets, such as divers. Traditional methods often fail to distinguish these targets from reverberation, and data-driven methods are constrained by the limited data on intruding targets. This paper introduces the online robust principal component analysis and multimodal anomaly detection (ORMAD) method to address these challenges. ORMAD efficiently performs online low-rank and sparse decomposition while utilizing unsupervised multimodal anomaly detection to enhance detection performance. The multimodal anomaly detection process involves two phases: modality extraction and anomaly detection. During modality extraction, echo data are separated into echo structure and spatial trajectory modalities, providing complementary information that improves the network representation of both reverberation and moving targets. The subsequent anomaly detection phase unsupervisedly learns the modalities of fluctuating reverberation, thereby achieving stable reconstruction while maintaining sensitivity to moving targets. This sensitivity allows effective identification of moving targets by detecting reconstruction loss. Experimental results demonstrate that ORMAD effectively improves detection performance in complex reverberation scenarios. In a real-world sonar dataset, ORMAD increased the average precision for detecting diver targets from 60% to 75% compared to the state-of-the-art method.

Study on Improving Detection Performance of Wildfire and Non-Fire Events Early Using Swin Transformer

Study on Improving Detection Performance of Wildfire and Non-Fire Events Early Using Swin Transformer

Waste drilling fluid flocculation identification method based on improved YOLOv8n.

Efficient identification of the flocculation state of waste drilling fluid remains a significant challenge. This study proposes an improved You Only Look Once version 8 nano-algorithm (YOLOv8n), specifically optimized for real-time monitoring of drilling fluid flocculation under field conditions. The algorithm employs MobileNetV3 as the backbone network to minimize memory usage, improve detection speed, and reduce computational requirements. The integration of the efficient multi-scale attention mechanism into the cross-stage partial fusion module effectively mitigates detail loss, resulting in improved detection performance for images with high similarity. The wise intersection over union loss function is employed to accelerate bounding box convergence and improve inference accuracy. Experimental results show that the enhanced YOLOv8n algorithm achieves an average recognition accuracy of 98.6% on the experimental dataset, a 4.8% improvement over the original model. In addition, the model size and parameter count are reduced to 2.9 MB and 2.8 Giga Floating-Point Operations Per Second (GFLOPS), respectively, compared to the original model, reflecting a reduction of 3.2 MB and 5.3 GFLOPS. As a result, the proposed flocculation recognition algorithm is highly deployable and effectively predicts flocculation state changes across varying working conditions.

Vacuum degree detection performance improvement of vacuum switches based on NELIBS

We aimed to investigate the enhancement effects of silver nanoparticles on signals under low-pressure. By analyzing spectral data, plasma images, and radiation integral intensity, we seek to improve the accuracy of vacuum level measurements.

BGI-YOLO: Background Image-Assisted Object Detection for Stationary Cameras

This paper proposes a method enhancing the accuracy of object detectors by utilizing background images for stationary camera systems. Object detection with stationary cameras is highly valuable across various applications, such as traffic control, crime prevention, and abnormal behavior detection. Deep learning-based object detectors, which are mainly used in such cases, are developed for general purposes and do not take advantage of stationary cameras at all. Previously, cascade-based object detection methods utilizing background have been studied for stationary camera systems. These methods typically consist of two stages: background subtraction followed by object classification. However, their object detection performance is highly dependent on the accuracy of the background subtraction results, and numerous parameters must be adjusted during background subtraction to adapt to varying conditions. This paper proposes an end-to-end object detection method named BGI-YOLO, which uses a background image simply by combining it with an input image before feeding it into the object detection network. In our experiments, the following five methods are compared: three candidate methods of combining input and background images, baseline YOLOv7, and a traditional cascade method. BGI-YOLO, which combines input and background images at image level, showed a detection performance (mAP) improvement compared to baseline YOLOv7, with an increase of 5.6%p on the WITHROBOT S1 dataset and 2.5%p on the LLVIP dataset. In terms of computational cost (GFLOPs), the proposed method showed a slight increase of 0.19% compared to baseline YOLOv7. The experimental results demonstrated that the proposed method is highly effective for improving detection accuracy without increasing computational cost.