Search

The global decrease in native pollinators poses a substantial challenge to agricultural production and food security, particularly in malnutrition prone countries. There is a huge potential in nanobots which nanobots can autonomously explore complicated surroundings to perform targeted pollination by utilizing Convolutional Neural Network (CNN) and You Only Look Once (YOLO) for real-time flower recognition and categorization. A comprehensive study has been depicted in the article and shows how various approaches have utilized Simultaneous Localization and Mapping (SLAM) to improve spatial awareness, allowing the nanobots to adjust directions on the go. The study also presents an extensive literature survey shows the most productive countries that have contributed to this field of study. Furthermore, this article also presents a broad array of studies that have used deep learning approaches to enhance sustainable agriculture methods. Additionally, the study contributes a comparative study of the results that have been showcased in the literature. A performance metric comparison was performed among the deep learning model utilized and for various optimizers used along with the Deep Learning models. The article concludes with potential limitations, challenges and opportunities for advancing sustainable agriculture using deep learning models.

Intracranial hemorrhages (ICH) are a significant challenge in emergency medicine due to the critical nature of a timely and accurate diagnosis. This study evaluates the performance of six versions of the You Only Look Once (YOLO) object detection model, from YOLOv5 to YOLOv10, in detecting ICH using computed tomography (CT) scans. The primary focus is understanding the advancements in YOLO architectures over time and their impact on detection accuracy and inference speed. The study used the Brain Hemorrhage Extended Dataset (BHX), comprising 491 CT scans with annotations for six types of hemorrhages: epidural, subdural, subarachnoid, intraparenchymal, intraventricular, and chronic hemorrhage, and introduces a new data set obtained from a major hospital in Chile. The models were trained using a combination of single-class and multi-class approaches to address class imbalance and were evaluated based on precision, recall, F1 score, and mean average precision (mAP). The models were evaluated in three distinct contexts: 1) a biased scenario where images of the same individual could appear in both training and testing sets, 2) a cross-validation setup ensuring the independence of images by separating the sets based on subjects, and 3) an external validation using one dataset for training and the Chilean dataset for testing, maintaining full independence between training and evaluation. The findings indicate that YOLOv8 and YOLOv10 demonstrate superior detection accuracy and inference efficiency performance, respectively, compared to previous versions. In particular, with image independence, YOLOv8 reached the highest average mAP for all classes, with a score of 0.4. This comparative analysis provides information on the effectiveness of architectural advances in YOLO models for medical applications and suggests directions for future improvements in ICH detection.

This research aims to improve the automatic identification of armed people in surveillance videos. We focus on people armed with pistols and revolvers. Furthermore, we use the YOLOv4 to detect people and weapons in each video frame. We developed a series of algorithms to create a dataset with the information extracted from the bounding boxes generated by YOLOv4 in real-time. Thereby, we initially developed six-armed people detectors (APD) based on six machine learning models: Random Forest Classifier (RFC-APD), Multilayer Perceptron (MLP-APD), Support Vector Machine (SVM-APD), Logistic Regression (LR-APD), Naive Bayes (NB-APD), and Gradient Boosting Classifier (GBC-APD). These models use 20 predictors to make their predictions. These predictors are computed from the bounding box coordinates of the detected people and weapons, their distances, and areas of intersection. Based on our results, the RFC-APD was the best-performing detector, with an accuracy of 95.59%, a recall of 94.51%, and an F1-score of 95.65%. In this work, we propose to create selectors for deciding which APD to use in each video frame (APD4F) to improve the detection results. Besides, we implemented two types of APD4Fs, one based on a Random Forest Classifier (RFC-APD4F) and another in a Multilayer Perceptron (MLP-APD4F). We developed 44 APD4Fs combining subsets of the six APDs. Both APD4F types outperformed most of the independent use of all six APDs. A multilayer perceptron-based APD4F, which combines an MLP-APD, a NB-APD, and a LR-APD, presented the best performance, achieving an accuracy of 95.84%, a recall of 99.28% and an F1 score of 96.07%.

Existing traffic sign detection algorithms suffer from high computational complexity and large parameter sizes, limiting their deployability. The YOLO-TS model integrates the Normalized Wasserstein Distance (NWD) with the Complete Intersection over Union (CIoU) loss function, significantly enhancing the detection of small traffic signs. The integration of StarBlock from StarNet into the C2f module forms the C2f-Star configuration, which simplifies the architecture. In addition, the Slimneck design paradigm is introduced into the neck network to further reduce computational demands while maintaining model accuracy. Subsequently, a Squeeze and Excitations Shared Detection Head (SESDH) is developed to integrate a squeezing and excitation attention mechanism. This design helps diminish the intricacy of the network architecture, increase attention on areas where traffic signs are presence and improve the ability of the model to indicate objects in complex surroundings. Experimental results on the CCTSDB traffic sign dataset reveal that the updated algorithm decreased the number of parameters by 35.33%, reduced the computation by 35.80% and shrunk the model size by 32.94% of the baseline YOLOv8n model while improving the mAP@50 by 1%, and the F1 score by 1.32%. Compared to other algorithms, here, a good balance between accuracy and lightweight design is achieved.

The efficient and accurate detection of cracks in concrete structures is critical for maintaining structural integrity and safety. This study compares two state-of-the-art convolutional neural network (CNN) models, Mask R-CNN and YOLOv8, for automated concrete crack detection, each model representing two mainstream approaches for object detection and instance segmentation: single-stage and two-stage approach. We evaluate both models on 1,203 concrete images with 7:2:1 training, testing, and validation split, and assess their accuracy and processing speed. Mask R-CNN achieves a mean Intersection over Union (IoU) of 96.5% with a minimum IoU of 77% and higher consistency, compared to YOLOv8’s 90.6%, which often shows complete failure with IoU of 0%. In terms of computation speed, YOLOv8 shows 0.3225 s of average processing time per image, slightly outperforming the speed of Mask R-CNN, 0.4867 s. Despite YOLOv8’s faster processing speed, considering the characteristics of concrete crack detection tasks where accuracy should be prioritized over speed, Mask R-CNN seems a more proper model for reliable crack detection. We also show the accuracy of Mask R-CNN for crack detection tasks can be further enhanced by employing the ResNeXt backbone.

Steel materials are extensively used across various industries. Detecting surface defects in steel strips during production processes is crucial. Existing steel surface defect detection methods exhibit inadequate accuracy and excessive computational complexity, posing challenges for real-time industrial deployment. In this paper, a novel model is designed named GDM-YOLO, specifically tailored for steel surface defect detection tasks, built upon the YOLOv8s network. Firstly, the Space-to-Depth Ghost Convolution (SPDG) downsampling module is introduced and used in the backbone network, aimed at minimizing information loss during downsampling operations while optimizing computational efficiency. Secondly, this work introduces the C2f-Dilated-Reparam-Block (C2f-DRB) module, leveraging reparameterization and large kernel convolutions to enhance feature extraction capabilities without compromising inference costs. Lastly, the novel Multiscale Feature Enhancement Block (MFEB) module was designed, to enhance the small target detection layer by integrating multi-scale feature fusion, further improving detection accuracy. Experimental results demonstrate a 3% improvement in detection accuracy on the NEU-DET dataset compared to the baseline YOLOv8s model. Our approach achieves superior detection performance while reducing parameter requirements and computational complexity, meeting the real-time demands of steel surface defect detection in industrial production.

Aiming at the YOLO (You Only Look Once) algorithm’s low detection accuracy caused by the complex background environment and large target scale difference in optical remote sensing image detection, the lightweight convolution fusion attention mechanism based AMH-YOLOv8 (Attention Mechanisms Hybrid- YOLOv8) target detection algorithm is proposed in this paper. In this algorithm, the BiFormer attention mechanism is added to enhance the detection performance for small targets. Effectively captures local and global information in remote sensing images, improving the accuracy and generalisation of target detection; secondly, using the lightweight convolution GSConv instead of the original normal convolution. Effectively reduces model size while ensuring that model performance is not compromised. And optimised computation and parameter count improvements due to BiFormer; finally, the SPPCSPC space pyramid structure was added. Effectively enhances the feature extraction capability of the model and reduces the probability of missed and false detections.In order to verify the effectiveness of the algorithm, experimental analyses were conducted on two publicly available datasets, namely DIOR and DOTA. Experimental results indicate that the AMH-YOLOv8 algorithm achieved an impressive detection accuracy of 87.6% and 72.9% in mAP@0.5, The results show that the algorithm in this paper has improved the effectiveness of target detection in optical remote sensing images. And it can better cope with the problems of complex background environment, dense distribution of small targets and large differences in target scales.

Skin defects significantly impact the quality of rose apples. The presence of damaging defects, such as holes, rot, or piercing marks in exported products can negatively impact customer trust. In addition, the products are further graded based on the total area occupied by acceptable defects such as scratches, rubbing, and shallow scars. This study investigated the application of instance segmentation network, specifically single-stage and two-stage approaches, for detecting the areas of rose apple skin and surface lesions. The surface defects were divided into two classes including acceptable and damaging defects. For improving networks capability, anchor boxes were optimized for detecting multi-size rose apple skin and defects. Additionally, the both instance segmentation networks were further adjusted their network size. The results revealed that optimizing anchor boxes based on training data distribution led to significant improvement in detection performance for both single-stage and two-stage networks, compared to K-means clustering approach. In contrast, adjusting the network size had a less pronounced impact. Finally, YOLOv5 with optimized anchor boxes significantly increased the F1-score of detecting defects by 23.98%. However, this trade-off with the lower skin detection F1-score only by 5.41%, compared to Mask R-CNN. Furthermore, YOLOv5 demonstrated greater robustness compared to Mask R-CNN in distinguishing between objects with visual characteristics that resemble defects, such as brightness, shadow, and skin texture. The outcomes provided strong evidence that YOLOv5 offers superior efficiency in detecting multi-scale objects, despite having a network size and inference time that are 1.89 and 1.49 times smaller, respectively, compared to Mask R-CNN.

This study researches the usage of YOLOv4 for real-time loose fruit detection in oil palm plantations as the first step in implementing automation in the collection of loose fruits. Our system leverages high-resolution video data (4K and 1080p) from various plantation settings. To address the challenges of detecting small and numerous loose fruits, we introduced an image preprocessing technique called “image tiling” into the vision system workflow. We studied the effects this has on the performance of the detection model. This involves slicing the image into smaller sections (i.e., tiles) for individual processing by YOLOv4 and YOLOv4-tiny models, enhancing detection accuracy. Refined models (YOLOv4-tiling and YOLOv4-tiny-tiling) are then evaluated. YOLOv4 achieved the highest precision (97%) and F1-score (86.3%), while YOLOv4-tiling offered a slight improvement in recall (80.8%). Notably, YOLOv4-tiny, initially underperforming (precision: 37.2%, recall: 20.9%, F1-score: 25%), showed significant improvement with tiling (precision: 90.5%, recall: 67.1%, F1-score: 73.8%). Also, replacing the SPP layer in YOLOv4 with SPP-Fast resulted in increased precision (92.6%) and a significantly improved F1 score of 91.4%. This vision system was then integrated with a custom-designed Loose Fruit Collector Robot through the Robot Operating System (ROS).

The limited accuracy of detecting small objects with current AI algorithm is becoming prominent under background of security check using low-resolved terahertz images. To solve this problem, a modified model based on YOLOv8 architecture – YOLOv8-SAB is built to achieve satisfactory detection of suspicious small objects. Specifically, the model integrates Shuffle attention mechanism into YOLOv8, capturing channel dependence and spatial pixel relationship through Channel Shuffle operation and attention mechanism, thus improving the detection effect of small objects. In addition, BiFPN (Bi-directional Feature Pyramid Network) is used in feature fusion to overcome low resolution of terahertz images, and employ SIoU (SCYLLA IoU) as a friendly optimization object. Experimental results indicate that the inference speed of YOLOv8-SAB reaches 1.7ms to meet the fast detection requirement in real task. YOLOv8-SAB not only shows great potential in small object detection, but also has excellent detection ability for large objects. According to the self-made data set, mAP0.5 and mAP0.5-0.95 for small objects reach 0.93 and 0.716, respectively. As a comparison, mAP0.5 and MAP0.5-0.95 for large objects reach 0.98 and 0.84, respectively. This method plays an important role in the designing and deployment of intelligent real-time terahertz object detection system for security check.

Deep learning-based object detection technology often relies on non-maximum suppression (NMS) algorithms to eliminate redundant detections. However, the conventional NMS algorithm struggles with distinguishing between overlapping and small objects due to its simple constraints. While Soft-NMS offers a slight improvement in object detection performance, it still falls short in addressing this challenge. Our proposed solution, adjustable-NMS, represents a significant advancement. While performing comparably to NMS and Soft-NMS on less dense images where objects are easily countable, adjustable-NMS excels in scenarios with higher object density or smaller objects. In such cases, it outperforms both NMS and Soft-NMS, showcasing notably superior object detection capabilities. On average, the improvement achieved with adjustable-NMS reaches an impressive 33.3%. This demonstrates adjustable-NMS’s efficacy in enhancing object detection accuracy, particularly in challenging environments characterized by dense scenes or diminutive objects.

With the rapid proliferation of drones across various domains, aerial target detection has become increasingly crucial. However, the targets in aerial images present challenges such as scale variation, small size, and density, leading to suboptimal performance of current detectors on aerial images. Based on the aforementioned challenges, we design an efficient aerial target detection algorithm called DMFF-YOLO. Specifically, to address the issues of small target size and scale variation, we design the DMFF neck structure, adding a small target detection head to tackle the small target size problem, using the DMC module to fuse different scale features for enriching detailed information, and employing the DSSFF module to construct a scale sequence space to solve the target scale variation problem. In the network backbone, we employ RFCBAMConv modules as downsampling layers, which interact with receptive-field features to mitigate the information disparity caused by positional changes and outperform traditional convolutional layers. Finally, we design the Soft-NMS-CIoU module to address the issue of suppressing adjacent boxes due to dense targets. On the VisDrone dataset, compared to the original algorithm, our method reduces the number of parameters by 31.1% while achieving an 11.7% improvement in mAP50. Extensive experiments on the VisDrone, DOTA, and UAVDT datasets demonstrate that the proposed algorithm performs well in aerial image detection tasks.

Road target detection is essential for enhancing vehicle safety, increasing operational efficiency, and optimizing user experience. It also forms a crucial part of autonomous driving and intelligent monitoring systems. However, current technologies face significant limitations in multi-level feature fusion and the accurate identification of key targets in complex data environments. To address these challenges, this paper proposes an innovative algorithmic model called BiFPN GAM SimC2f-YOLO (BGS-YOLO), aimed at improving detection performance. Initially, this paper employs the Bidirectional Feature Pyramid Network (BiFPN) to effectively integrate multi-level features. This integration overcomes the limitations in feature extraction and recognition found in existing target detection algorithms. Following this, this paper introduces the Global Attention Module (GAM), which markedly improves the efficiency and accuracy of extracting key target information in complex data environments. Additionally, this paper innovatively designs the SimAM-C2f (SimC2f) network, further advancing feature expressiveness and fusion efficiency. Experiments on the public COCO dataset demonstrate that the BGS-YOLO model significantly outperforms the existing YOLOv8n model. Notably, it shows a 7.3% increase in mean average precision (mAP) and a 2.4% improvement in accuracy. These results highlight the model’s high precision and swift response in detecting road targets in complex traffic scenarios. Consequently, the BGS-YOLO model has the potential to significantly enhance road safety and contribute to a considerable reduction in traffic accident rates.

Target detection via remote sensing is extensively utilized across diverse domains because of its inherent potential value in applications. However, most objects within remote sensing images consist of multi-scale and dense small objects, observed from diverse angles against complex backgrounds, resulting in insufficient detection performance. To enhance the detection accuracy and robustness in detecting multi-scale objects, we present the YOLO-GE algorithm based on you only look once (YOLO). We introduce the ghost convolution hierarchical graph (G-HG) block that combines ghost convolutions and the cross-stage partial (CSP) strategy. This enhancement can efficiently utilize redundant feature maps, broaden the receptive field, and accurately extract multi-scale objects and advanced semantic features in complex backgrounds. By incorporating the G-HG block, we establish the ghost-convolution enhanced hierarchical graph (GE-HGNet) feature extraction backbone, thereby enhancing its ability to capture multi-scale object features and advanced semantic information. Additionally, we develop the E-SimAM attention mechanism using residual techniques to address the low-resolution feature loss limitation, thereby enhancing the precision in identifying low-resolution features against intricate backgrounds. Furthermore, to improve the capability of detecting densely packed small objects, we reconstruct the structure of the neck and add a tiny detection head. This additional tiny detection head is specifically designed to better focus on densely packed small targets, fully leveraging the fine-grained information in shallow feature maps. Extensive experiments conducted on the DIOR, NWPU VHR-10, and VisDrone2019 datasets demonstrate the effectiveness and robustness of our YOLO-GE algorithm. Notably, compared to the state-of-the-art algorithm, our YOLO-GE-n achieves improvements of 20.1% and 22.2% in mAP0.5 and mAP0.5:0.95 respectively on the VisDrone2019 dataset.

The hippocampal region is one of the most affected brain areas observed as a landmark in Magnetic Resonance Imaging (MRI) images for Alzheimer’s disease (AD) diagnosis. The diminished alterations in the hippocampal and degeneration of cholinergic circuits have been conclusively correlated with a decline in memory and cognitive function. However, the hippocampal region may not appear as clearly defined as other brain regions, making it difficult for neurologists and researchers to identify by visual inspection. The application of deep learning models to pinpoint the hippocampal region was initially valued. We assessed the ability of a deep learning model, You Only Live Once (YOLO), to detect hippocampal regions in three MRI image views and categories. The Alzheimer’s Disease Neuroimaging Initiative-first (ADNI−1) dataset was used with 220 subjects in three categories using the three YOLO models. We obtained the YOLO performance for hippocampal region detection with accuracy in three views and categories. The average mean Average Precision (mAP) performance accuracy for YOLOv3 was 0.87, YOLOv4 was 0.85, and YOLOv5 was 0.96, respectively. The high accuracy of the detection of the hippocampal region was remarkable. We found that the sagittal view was higher than the axial and coronal views. Simultaneously, the Mild Cognitive Impairment (MCI) in the coronal view was lower among the three models. The results showed that YOLOv5 is a suitable model for detecting the hippocampal region in MRI images, and the sagittal view is the most reliable for detecting the hippocampal region in diagnosing AD. Our findings demonstrate the importance of detecting the hippocampal region to diagnose AD and accurately analyzing the hippocampal area within the region. The YOLOv5 model substantially affected performance metrics and interpretability across the three views and categories.

The secure and stable operation of power transmission lines is essential for electrical systems. Given that abnormal targets such as bird’s nests and defective insulators may lead to transmission failures, timely detection of these targets is imperative. This paper introduces the YOLO-PowerLite model, an advanced lightweight object detection model based on YOLOv8n, designed for efficient, real-time detection on resource-constrained unmanned aerial vehicles (UAVs) equipped with edge computing platforms. In the feature fusion module, YOLO-PowerLite incorporates the innovative C2f_AK module, significantly reducing the number of parameters and enhancing the adaptability and fusion capability of features at different scales. Meanwhile, the adoption of the Bidirectional Feature Pyramid Network (BiFPN) further optimizes the efficiency and effectiveness of feature processing. In addition, the newly designed lightweight detection head significantly reduces the number of parameters and computational requirements. The integration of the Coordinate Attention mechanism in the backbone network enhances the model’s ability to focus on and recognize abnormal targets in complex backgrounds. Experimental results show that YOLO-PowerLite achieves a mAP@0.5 of 94.2%, maintaining the accuracy of the original YOLOv8n while significantly reducing parameters, FLOPs, and model size by 42.3%, 30.9%, and 40.4%, respectively. Comparative analysis shows that YOLO-PowerLite surpasses other mainstream lightweight models in detection accuracy and computational efficiency. Deployment on the NVIDIA Jetson Xavier NX platform demonstrates an average processing time of 31.2 milliseconds per frame, highlighting its potential for real-time applications in monitoring transmission lines.

The task of detecting surface defects on substation equipment faces several challenges, including a variety of target categories, the scarcity of original defect image data, complex environmental conditions, low accuracy in existing algorithms, as well as notable issues with false alarms and missed detections. Overcoming these obstacles is crucial for the successful implementation of intelligent inspection systems for substations. To address the problem of limited original data, we first employ the method of ADD-GAN to augment the image training set. Furthermore, this paper proposes a target detection model called YOLO-SD to detect various equipment defects in complex real-world scenarios. In order to enhance the network’s feature extraction capabilities in the presence of complex backgrounds and to improve detection accuracy, a novel deep perceptual feature extraction module named C3+ was introduced in this research. Furthermore, we incorporates SimAM into the neck network of YOLO-SD. This integration not only bolsters the network’s learning capacity but also equips it with the capability to autonomously learn and dynamically fine-tune attention weights to suit different input scenarios. To tackle the challenges posed by variations in size and shapes of different substation equipment defects in images, a novel fusion loss function NWD-CIoU is designed. The improvements enhance the accuracy and robustness of YOLO-SD in defect target detection across different scales. The experiment demonstrated that the YOLO-SD model achieved an mAP@0.5 of 90.3% and mAP@0.5:0.95 of 63.9% in detecting defects in substation equipment. The F1 score reached 81.1%, IoU value was 90.5%. This model realized accurate detection of multi-scale substation defect targets, reaching the state-of-the-art level in substation defects detection.

During the production process of steel plate, surface defect detection is crucial for high-quality products. For the existing defect detection method based on machine vision, there are various types of problems, such as large model calculations, low detection accuracy and difficulties of recognizing small defect targets. To reduce and solve these issues, the paper proposes a new defect detection model, simplified kernel and squeeze on a you only look once network (SKS-YOLO), which can achieve rapid and effective defect detection on steel plate. Firstly, it adopts EfficientNetv2 as the backbone, significantly reducing model calculations and accelerating training speed while maintaining accuracy. Subsequently, the atrous spatial pyramid pooling (ASPP) module is utilized to obtain a larger receptive field, extracting more feature information from surface defects. The integration of the squeeze excitation network (SE-Net) attention mechanism enhances capabilities of feature extraction furtherly. Then, the K-means algorithm is applied to cluster and obtain more suitable anchor frames for defect targets. It not only increases the number of positive samples, but also expedites model convergence. Finally, the loss function of simplified intersection over union (SIoU) is used to enhance the ability of model to locate and detect surface defect targets. The experimental results show that the mean average precision (mAP) is 89.40% at a detection speed of 55 frames per second (FPS), which is better than the state-of-the-art (SOTA) detection models.

Traditional palm oil production methods for evaluating fruit bunches (FFBs) are inefficient, costly, and have limited coverage. This study evaluates the performance of various YOLO models and other state-of-the-art object detection models using a novel dataset of oil palm fresh fruit bunches in plantations, captured in the plantation regions of Central Kalimantan Province, Indonesia. The dataset includes five ripeness classes (abnormal, ripe, underripe, unripe, and flower) and presents challenges such as partially visible objects, low contrast scenes, occluded and small objects, and blurry images. The proposed YOLOv8s Depthwise model was compared with other YOLO models, including YOLOv6s, YOLOv6l, YOLOv7 Tiny, YOLOv7l, YOLOv8s, and YOLOv8l. YOLOv8s Depthwise demonstrated a balanced performance, with a compact size (10.6 MB), fast inference time (0.027 seconds), and strong detection accuracy (mAP50 at 0.75, mAP50-95 at 0.481). Its rapid convergence and low training loss highlighted its efficiency, completing training in the shortest time of 2 hours, 18 minutes, and 30 seconds. Furthermore, it achieved low Mean Absolute Error (MAE) of 0.164 and Root Mean Square Error (RMSE) of 0.4, indicating precise counting capability. Hyperparameter tuning revealed that the YOLOv8s Depthwise model achieved optimal performance using the SGD optimizer with a batch size of 16 and a learning rate of 0.001, showing the best balance between accuracy and training efficiency. Data augmentation positively impacted model performance, resulting in improved performance metrics across various models. When evaluated against other state-of-the-art models on the same dataset, including Faster RCNN, SSD MobileNetV2, YOLOv4, and YOLOv9, YOLOv8s Depthwise surpassed other state-of-the-art models, including Faster R-CNN, SSD MobileNetV2, YOLOv4, and EfficientDet-D0 from previous research, in terms of speed, accuracy, and efficiency, making it ideal for real-time palm oil harvesting applications.

Deep learning has evolved dramatically in recent years, resulting in a wide range of useful applications in various fields, such as computer vision, natural language processing, robotics, and autonomous vehicles. This expansion demonstrates the high transformative potential of deep learning techniques in real-world applications. This study addresses two critical challenges in this context. First, we recognized the unique potential of hand wrist creases as robust biometric identifiers for individuals. Second, we acknowledge the importance of custom deep learning models in enhancing the accuracy of object detection for hand wrist creases. In response, we propose a novel deep learning method for hand wrist crease object detection by leveraging, custom yolov8 models in this research. Our experiments involved comparing the performance of the custom Yolov8 with the Yolov5 and Yolov8 models. Notably, the custom Yolov8x model achieved an impressive object-detection accuracy of 97.1%.This innovative approach has the potential to significantly affect the development of biometric security systems, particularly in scenarios where traditional biometric identifiers are impractical or unreliable. By harnessing the power of deep learning and custom model architectures, this study paves the way for more accurate and reliable biometric identification systems, contributing to the advancement of security measures in a wide range of applications and domains.

This paper deploys and explores variants of TinyissimoYOLO, a highly flexible and fully quantized ultra-lightweight object detection network designed for edge systems with a power envelope of a few milliwatts. With experimental measurements, we present a comprehensive characterization of the network’s detection performance, exploring the impact of various parameters, including input resolution, number of object classes, and hidden layer adjustments. We deploy variants of TinyissimoYOLO on state-of-the-art ultra-low-power extreme edge platforms, presenting a detailed comparison on latency, energy efficiency, and their ability to efficiently parallelize the workload. In particular, the paper presents a comparison between a RISC-V-based parallel processor (GAP9 from GreenWaves Technologies) with and without use of its on-chip hardware accelerator, an ARM Cortex-M7 core (STM32H7 from ST Microelectronics), two ARM Cortex-M4 cores (STM32L4 from ST Microelectronics and Apollo4b from Ambiq), and a multi-core platform aimed at edge AI applications with a CNN hardware accelerator (MAX78000 from Analog Devices). Experimental results show that the GAP9’s hardware accelerator achieves the lowest inference latency and energy at $\mathrm {2.12ms }$ and $\mathrm {150~\mu \text {J} }$ respectively, which is around 2x faster and 20% more energy efficient than the next best platform, the MAX78000. The hardware accelerator of GAP9 can even run an increased resolution version of TinyissimoYOLO with $112\times 112$ pixels and 10 detection classes within 3.2 ms, consuming $\mathrm {245~\mu \text {J} }$ . We also deployed and profiled a multi-core implementation on GAP9 at different core voltages and frequencies, achieving $\mathrm {11.3ms }$ with the lowest-latency and $\mathrm {490~\mu \text {J} }$ with the most energy-efficient configuration. With this paper, we demonstrate the flexibility of TinyissimoYOLO and prove its detection accuracy on a widely used detection dataset. Furthermore, we demonstrate its suitability for real-time ultra-low-power edge inference.

The leap forward in research progress in real-time object detection and classification has been dramatically boosted by including Embedded Artificial Intelligence (EAI) and Deep Learning (DL). Real-time object detection and classification with deep learning require many resources and computational power, which makes it more difficult to use deep learning methods on edge devices. This paper proposed a new, highly efficient Field Programmable Gate Array (FPGA) based real-time object detection and classification system using You Only Look Once (YOLO) v3 Tiny for edge computing. However, the proposed system has been instantiated with Advanced Driving Assistance Systems (ADAS) for evaluation. Traffic light detection and classification are crucial in ADAS to ensure drivers’ safety. The proposed system used a camera connected to the Kria KV260 FPGA development board to detect and classify the traffic light. Bosch Small Traffic Light Dataset (BSTLD) has been used to train the YOLO model, and Xilinx Vitis AI has been used to quantify and compile the YOLO model. The proposed system can detect and classify traffic light signals from a high-definition (HD) video streaming in 15 frames per second (FPS) with 99% accuracy. In addition, it consumes only 3.5W power, demonstrating the ability to work on edge devices. The on-road experimental results represent fast, precise, and reliable detection and classification of traffic lights in the proposed system. Overall, this paper demonstrates a low-cost and highly efficient FPGA-based system for real-time object detection and classification.

Cervical spine fractures are a medical emergency that can cause permanent paralysis and even death. Traditional fracture detection techniques, such as manual radiography image interpretation, are time-consuming and prone to human error. Deep learning algorithms have shown promising results in various medical imaging applications i.e., disease diagnosis, including fracture detection of bones. In this study, we propose a two-stage approach for detecting cervical spine fractures. The first stage employs a convolutional neural network (CNN) model to determine the presence or absence of a fracture in the cervical spine, using a dataset of cervical spine Computed Tomography (CT) scan images as well as Grad-CAM for enhanced visualization and interpretation. In the second stage, our focus shifts to specific vertebrae within the cervical spine. To accomplish this task, we trained and evaluated the performance of the YOLOv5 and YOLOv8 models with 9170 images consisting of seven vertebrae. The detection results of both YOLO versions are compared and evaluated. The precision, recall, mAP50, and mAP50-90 were 0.900, 0.890, 0.935, 0.872, respectively. The results of this research demonstrate the potential of deep learning-based approaches for cervical spine fracture detection. By automating the detection process, these algorithms can assist radiologists and healthcare professionals in making accurate and timely diagnoses, leading to improved patient outcomes.

To solve the problems of limited computing power resources, low accuracy of small target detection, high miss rate, and poor real-time detection of mobile vehicle platforms in the automatic driving environment, The present study introduced a one-stage target detection algorithm TTD-YOLO (Traffic Target Detection YOLO) that improved YOLOV5-S, which is enhanced in four aspects: Enhanced the network’s multi-scale feature extraction performance through the utilization of the improved M-ELAN architecture; added 3D attention mechanism SimAM to the network structure to enable the network to learn important feature information and enhance the efficiency of detecting accuracy; the parameter ratio of backbone and neck is adjusted to close to 1:1 by adjusting the number of output channels and stacking times of CSPLayer modules in the backbone and neck, while maintaining the model complexity, experiments show that improving the neck’s parameter ratio helps enhance the efficiency of detecting accuracy without changing the network structure; used EIoU loss instead of the bounding box loss function accelerates network convergence and improved detection accuracy. Under the condition of avoiding significantly changing the network structure, our TTD-YOLO outperforms the baseline model and other mainstream object detection algorithms such as Faster RCNN, SSD, and YOLOX-S on the autopilot dataset SODA10M with fewer parameters, higher detection accuracy, and faster inference speed. Compared to the baseline model, the model parameters decreased by 8.6%, average precision(mAP@.5:.95)increased by 2.5%, and the inference speed under the same experimental platform increased by 4.8%.

Accurate real-time observations of wheat head growth are crucial for effective agricultural management. However, the dense distribution of wheat heads often leads to severe overlap in imagery. Existing target detection algorithms face challenges in overcoming this problem, rendering them ineffective for real-time field computations using portable devices. Therefore, this study proposes a lightweight you-only-look-once (YOLO) model with a simplified structure and a more powerful attention mechanism. A limitation of the traditional YOLO model is its complex structure, as it requires a substantial number of parameters and its accuracy is unsatisfactory. We remove the modules designed for large targets and reduce the number of detection heads from three to two. Moreover, we add an improved feature pyramid network to the neck, resulting in improved parameter count and accuracy over traditional YOLO methods. To improve inferencing, we replace the spatial pyramid pooling (SPP) module with a simplified SPP-fast type. Finally, a large separable kernel attention and wise intersection-over-union method are introduced to integrate the attention mechanisms, and we replace the loss function to improve the discriminative capabilities of the model. Experimental results on the Global Wheat Head Dataset demonstrate a 53% reduction in memory usage, 27% decrease in computational load, and 5.2 frames-per-second increase in detection speed over extant methods. The proposed model also achieves 3.9%, 2.1%, and 1.3% improvements in terms of precision, recall, and mean average precision, respectively, and it exhibits lightweight and portable characteristics. Overall, the proposed method has fast inference speed and effectively alleviates the problem of low recognition of dense wheat head.

As an important raw material supporting the development of society, wood lumber is widely used in the construction and furniture industries. However, traditional methods for detecting surface defects in wood face the challenges such as poor recognition, low efficiency and narrow applicability. To tackle these challenges, this paper proposes a Wood Lumber Surface-defect Detection-YOLO (WLSD-YOLO) model for the detection of surface defects in wood lumber. Firstly, this model introduces the squeeze-and-excitation (SE) attention mechanism in the backbone feature network component, which enhances the ability of capturing defects. Furthermore, a new GVC-neck layer structure is proposed to reduce the number of parameters and improve the accuracy of detection. Lastly, the combination of Normalized Weighted Distance Loss (NWD) small target detection algorithm and the Wise Intersection Over Union (WIOU) loss function is used to replace the original loss function to enhance the small target detection capability. The experimental results show that WLSD-YOLO achieves an average recognition accuracy of 76.5% for wood lumber defects. Compared with the original model YOLOv8, the mean average precision(mAP) is improved by 2.9% and the frames per second (FPS) is improved by 3.8. Meanwhile, WLSD-YOLO reduces the number of parameters to better detect several specific defects that are difficult to identify, which provides high application value for wood lumber processing and manufacturing industry.

Deep learning has revolutionized object detection, with YOLO (You Only Look Once) leading in real-time accuracy. However, detecting moving objects in visual streams presents distinct challenges. This paper proposes a refined YOLOv8 object detection model, emphasizing motion-specific detections in varied visual contexts. Through tailored preprocessing and architectural adjustments, we heighten the model’s sensitivity to object movements. Rigorous testing against KITTI, LASIESTA, PESMOD, and MOCS benchmark datasets revealed that the modified YOLOv8 outperforms the state-of-the-art detection models, especially in environments with significant movement. Specifically, our model achieved an accuracy of 90%, a mean Average Precision (mAP) of 90%, and maintained a processing speed of 30 frames per second (FPS), with an Intersection over Union (IoU) score of 80%. This paper offers a detailed insight into object trajectories, proving invaluable in areas like security, traffic management, and film analysis where motion understanding is critical. As the importance of dynamic scene interpretation grows in artificial intelligence and computer vision, the proposed enhanced YOLOv8 detection model highlights the potential of specialized object detection and underscores the significance of our findings in the evolving field of object detection.

Currently, defect detection for photovoltaic (PV) electroluminescence (EL) images faces three challenges: limited training data and complex backgrounds result in low accuracy in detecting defects; the diverse shapes of specific defects often lead to frequent false alarms; and existing models still require improvement in accurately recognizing these 12 specific defects. An innovative detector, called the Reconfigurable Adaptive Focus Background Suppression Detector (RAFBSD), is proposed to tackle these challenges. This approach utilizes the advanced Deformable Attention Transformer (DAT) technology to effectively focus on regions with defects. By doing so, it captures more informative features while significantly reducing noise interference. Additionally, the Omni-dimensional Dynamic Convolution (ODConv) and the Receptive-Field Attention convolutional (RFAConv) modules enhance accuracy by effectively allocating attention weights to optimize model performance. Compared to YOLOv8, RAFBSD has demonstrated substantial enhancements in both mAP50 (by 4.7%) and mAP50-95 (by 5.3%). Furthermore, RAFBSD demonstrates an outstanding capability to precisely identify flaws in 12 different classifications of PVEL images, exceeding the recognition capacity of YOLOv8 models by three additional types. These findings unequivocally establish RAFBSD as an efficient and groundbreaking object detection model.

Intracranial hemorrhage (ICH) is an emergency and a potentially life-threatening condition. Automated segmentation of ICH from head CT images can provide clinicians with volumetric measures that can be used for diagnosis and decision support for treatment procedures. Existing solutions typically involve training deep learning models to perform segmentation directly on the whole CT image. However, datasets with segmentation masks are typically very small in comparison with datasets with bounding boxes. Thus, we propose a two-stage approach that utilizes both bounding boxes and segmentation masks to help improve segmentation performance. In the first stage, ICH regions are detected and localized with bounding boxes surrounding the lesion by using a supervised YOLOv5 object detector. In the second stage, the localized ICH foreground is automatically segmented using TransDeepLab, an attention-based transformer network. Although we utilize both ground-truth bounding boxes and segmentation masks, different datasets can be used to train each stage. There is no requirement for pairing up bounding boxes and segmentation masks to train the model. Since bounding box annotations are available in larger quantities than segmentation masks, our approach allows these large datasets of bounding boxes to be used to improve ICH segmentation performance. On our dataset of segmentation masks, we demonstrated that our proposed two-stage YOLOv5+ TransDeepLab model outperformed segmentation methods such as SegResNet by 8% in terms of Dice score. Given ground truth bounding boxes, a Dice score of 0.769 is achieved, outperforming state-of-the-art methods such as nnU-Net. In sum, our proposed two-stage approach produces more accurate binary segmentation of ICH for neuroradiologists and these improved measurements could potentially aid their clinical decision-making process.

YOLO (You Only Look Once) is an extensively utilized object detection algorithm that has found applications in various medical object detection tasks. This has been accompanied by the emergence of numerous novel variants in recent years, such as YOLOv7 and YOLOv8. This study encompasses a systematic exploration of the PubMed database to identify peer-reviewed articles published between 2018 and 2023. The search procedure found 124 relevant studies that employed YOLO for diverse tasks including lesion detection, skin lesion classification, retinal abnormality identification, cardiac abnormality detection, brain tumor segmentation, and personal protective equipment detection. The findings demonstrated the effectiveness of YOLO in outperforming alternative existing methods for these tasks. However, the review also unveiled certain limitations, such as well-balanced and annotated datasets, and the high computational demands. To conclude, the review highlights the identified research gaps and proposes future directions for leveraging the potential of YOLO for medical object detection.

In today’s world, deepfake technology is being used to generate fake images, sounds, and videos from real images and sounds using deep learning and artificial intelligence techniques. It is possible to manipulate medical images with this technology. The manipulation of medical images can lead to incorrect diagnoses by medical professionals, disrupting the functioning of hospitals. As a result of these disruptions, hospitals may experience significant financial and life-threatening problems. In this study, it is aimed to obtain an effective deep learning-based method to detect manipulated medical images. Initially, two distinct datasets are created which contain Knee Osteoarthritis X-ray and lung CT scans. Data pre-processing and augmentation methods are applied for data standardization and variation. The instances in datasets are labeled as real or fake. The medical deepfake distinguish ability of YoloV3, YoloV5nu, YoloV5su, YoloV8n, YoloV8s, YoloV8m, YoloV8l, YoloV8x models tested on these datasets. In the analysis performed, all YOLO models showed almost full success in distinguishing Knee Osteoarthritis X-ray images. In lung CT scan images, although YoloV8 models generally achieved good performance, the YoloV5 models gave the best and worst results. While the best result was obtained from YoloV5su with a recall value of 0.997, the worst result was obtained from the YoloV5nu model with a recall value of 0.91. Furthermore, the best model (YoloV5su) works 60% faster than YoloV8x model, which has the second highest performance. The findings show that YoloV5su can be used for fast and accurate medical deepfake detection.

Aiming at the problems of tiny targets, large target scale changes, and background information interference in target detection of UAV(Unmanned Aerial Vehicle) aerial images, a revised UAV target detection algorithm MCA-YOLOv7 based on YOLOv7 is proposed, and the algorithm advances from the following points: optimizing the FPN(Feature Pyramid Networks) structure to increase the small-target detection layer, and boosting the network’s detection ability for small targets. To enhance the multi-scale feature extraction capability, the Efficient Multi-Scale Attention(EMA) is added. In order to reduce the complexity of the model and reduce the confusion of background information, the context aggregation block (CABlock) was introduced and improved, and an effective context aggregation block (ECABlock) was proposed. The loss function CIoU is enhanced and a new loss function FCIoU is proposed, which accelerates the convergence speed of the model, and obtains more accurate regression results. The experimental results demonstrate that the MCA-YOLOv7 model reduces the number of model parameters by 4.7 M and increases the average accuracy (mAP@0.5) by 2.9% when compared to YOLOv7 on the VisDrone2019 dataset. The new algorithm is more capable of handling situations involving UAV aerial photography.

This paper implements a systematic methodological approach to review the evolution of YOLO variants. Each variant is dissected by examining its internal architectural composition, providing a thorough understanding of its structural components. Subsequently, the review highlights key architectural innovations introduced in each variant, shedding light on the incremental refinements. The review includes benchmarked performance metrics, offering a quantitative measure of each variant’s capabilities. The paper further presents the performance of YOLO variants across a diverse range of domains, manifesting their real-world impact. This structured approach ensures a comprehensive examination of YOLOs journey, methodically communicating its internal advancements and benchmarked performance before delving into domain applications. It is envisioned, the incorporation of concepts such as federated learning can introduce a collaborative training paradigm, where YOLO models benefit from training across multiple edge devices, enhancing privacy, adaptability, and generalisation.

An improved YOLOv8 detection method is proposed for detecting distracted driving behavior and driver’s emotion. Unlike the commonly used YOLOv8 method, an attention mechanism named MHSA and a CNN module are synthesized to ensure improved performance in terms of accuracy and convergence, where MHSA is used to detect distracted driving behavior and CNN is used to detect driver’s emotion. The FER2013 dataset and collected dataset are used to train the improved YOLOv8. The training results show that the proposed YOLOv8 demonstrates improved performance compared with the commonly used YOLO based methods. Finally, the validity of the proposed YOLOv8 method is illustrated through implementations in Jetson Nano platform, where the TensorRT and DeepStream methods in the Jetson Nano device are used to optimize the volume and operational speed of the proposed YOLOv8 method, respectively. Test results show that the proposed YOLOv8 method can yield better real-time and accuracy properties.

Accurate, reliable, and fast intelligent detection of leather surface defects has become an important subject in industrial inspection, which aims at improving production efficiency and increasing automation levels. This work focuses on the rapid defect recognition and localization of leather surface defects for industrious applications, which is based on the state-of-the-art real-time detection model YOLO. Three experimental Schemes with different challenges were designed to find the optimal YOLO-based leather surface defect detection scheme. Typical tanned leather surface defect images from the factory were collected, which are comprised of eight types of defects, namely rotten surface, hole, scratch, crease, healed injury, bacterial injury, growth line, and pinhole, which exhibit variations in shapes, sizes, and colors, reflecting the various characteristics found in tanned leather defects. A comprehensive and in-depth review of the YOLO series of models is presented, including YOLOv1 to YOLOv8. The systematic and extensive experiments were conducted,which indicate that the YOLO models can simultaneously detect multiple types of defects present in each leather image. The multi-defect detection task achieved a maximum of 52.3% mean average precision (mAP), 58.2% precision, and 68.7% recall. For single-class detection tasks, the highest performance reached 85.1% mAP, 90.9% precision, and 81.8% recall. These works provide feasible intelligent solutions for surface defects in the leather industry, laying a solid foundation for the design and development of new solutions for leather defect detection.

Currently, with the widespread application of embedded technology and the continuous improvement of computational power in mobile terminals, the efficient deployment of algorithms on embedded devices, while maintaining high accuracy and minimizing model size, has become a research hotspot. This paper addresses the challenges of deploying the YOLOv8 algorithm on embedded devices and proposes a novel lightweight object detection algorithm focusing on small object detection. We optimize the model through two key strategies, aiming to achieve lightweight deployment and improve the accuracy of small object detection. Firstly, GhostNet is introduced as the backbone network for YOLOv8 in order to achieve lightweight deployment. By using some cost effective operations to generate redundant feature maps, we not only reduce the number of model parameters while ensuring better detection results, but also improve the speed of the model. Secondly, a new multi-scale attention module is designed to enhance the network’s acquisition of crucial information for small targets, which includes a multi-scale fusion attention mechanism and the Soft-NMS algorithm. The multi-scale fusion attention mechanism captures key features of discriminative small targets in the feature map tensor from both spatial and channel dimensions, suppressing non-key information, reducing the impact of complex and unimportant information in the image, enhancing the network model’s learning ability for important features of small targets. The Soft-NMS method improves accuracy by significantly reduces false positives in the detection results. To validate the performance of our proposed method, we conducted validation experiments on the PASCAL VOC dataset and evaluated the model’s generalization ability on the MS COCO dataset. The experiments results demonstrate that our model achieves a significant improvement in small object detection, with a 5.41% increase in detection accuracy compared to the existing YOLOv8. Meanwhile, FLOPs are reduced by 49.62%, and the number of model parameters is reduced by 48.66%. These results fully confirm the effectiveness of our innovative method in achieving both lightweight deployment and significant efficacy in small object detection tasks.

Since the UAV is far away from the detected target when performing target detection at high altitudes, there is a significant difference in the size of the detected object, and there are problems such as the detected target being blocked by the object. This paper proposed an improved algorithm BDH-YOLO for UAV high-altitude small target detection based on YOLOv8s. The backbone network of YOLOv8s is used, and the weighted bidirectional feature pyramid network (BiFPN) is combined with the backbone network to enhance the integration of multi-level features and obtain more multi-scale semantic information. Deal with multiple downsampling of the feature map will lead to the reduction of the resolution of the feature map, which will make the pixels on the feature map overlap, thus losing a large amount of spatial information, a dynamic detection head DyHead (Dynamic Head) combined with self-attention is introduced, which significantly improves the performance of the object detector by combining the perception capabilities of the three dimensions of scale perception, spatial perception, and task perception. Experimental results show that on the VisDrone2019-DET dataset, Compared with the YOLOv8s model, the BDH-YOLO model has an average accuracy (mAP@0.5) improvement of 3 targets than other mainstream models.

Leveraging the excellent feature representation capabilities of neural networks, deep learning methods have been widely adopted for object detection in synthetic aperture radar (SAR) images. However, persistent challenges are encountered in SAR ship detection due to factors such as small ship sizes, high noise levels, multiple targets, and scale variations. To address these complexities, in this paper, the MSFA-YOLO algorithm, a novel multiscale SAR ship detection approach em-powered by a fused attention mechanism, is presented. The proposed algorithm incorporates several key enhancements. The fused attention c2fSE module is integrated into the YOLOv8n baseline network to optimize feature extraction for SAR ships. In addition, the DenseASPP module is incorporated to enhance the model’s adaptability to ships of varying scales, improving its ca-pability to accommodate larger ships within lower model scales. Furthermore, the Wise-IoU loss function is adopted, and a dynamic non-monotonic focusing mechanism is employed for bounding box loss, significantly enhancing the model’s ability to handle low-quality images. Extensive experiments conducted on benchmark datasets, namely SAR-Ship-Dataset, SSDD, and HRSID, validate the robustness and reliability of the proposed model. Experimental results demonstrate significant performance improvements over YOLOv8n: a 3.1% enhancement in mAP75 and a 2.1% boost in mAP50–95 on the SAR-Ship-Dataset, a 0.7% increase in mAP75 and a 0.5% increase in mAP50–95 on the SSDD dataset, and a 1.8% increase in mAP75 and a 0.7% increase in mAP50–95 on the HRSID dataset. Exhibiting superior performance to existing SAR ship detection models in terms of accuracy, the MSFA-YOLO algorithm represents a significant advancement, establishing itself as the current state-of-the-art algorithm in SAR ship detection.

Recent advancements in image processing and computer vision have brought significant transformations in healthcare technology, leading to significant improvements in diagnosis accuracy, cost-effectiveness, and time efficiency. Magnetic Resonance Imaging (MRI) is employed by the radiologist for its remarkable ability to detect even the most subtle brain abnormalities. This study considers a comprehensive analysis of the two prominent object identification frameworks, YOLOv5 and YOLOv7, leveraging state-of-the-art deep learning architectures to classify and detect brain cancers within MRI. The brain tumor dataset encompasses three distinct classes, including meningiomas, gliomas and pituitary tumors. To ensure precise segmentation of the tumor regions, the preprocessing phase incorporates advanced mask alignment techniques. This preprocessed dataset has been used to evaluate the performance of the deep learning models for brain tumor detection and classification. From the numerical results of YOLOv5, it was noticed that a recall score of 0.905 for box detection and 0.906 for mask segmentation, with a precision score of 0.94 and 0.936, respectively. At an IoU threshold of 0.5, both box detection and mask segmentation achieve a mAP of 0.947, whereas, at an IoU threshold of 0.5 to 0.95, they achieve mAPs of 0.666 and 0.657, respectively. In comparison, YOLOv7 exhibits strong performance with box detection accuracy of 0.936 and a mask segmentation accuracy of 0.935. The recall score are 0.904 for box detection and mask segmentation is 0.903. Notably, the mAP result at the IoU threshold of 0.5 are 0.94 for box detection and mask segmentation is 0.941. Over the broader IoU spectrum of 0.5 to 0.95, the mAP was 0.677 for box detection and 0.659 for mask segmentation. To underscore the novelty of the approach, the performance of the proposed framework is systematically compared with established methods such as RCNN, Faster RCNN, and Mask RCNN.

Insulators play a pivotal role in power transmission lines, and the timely detection of defects in insulators is crucial to prevent potentially catastrophic consequences in terms of human lives and property. This paper proposes an insulator defect detection algorithm, named Insulator Lack-You Only Look Once (IL-YOLO), addressing the limitations observed in existing research concerning the complex background and multi-target challenges in insulator detection. The IL-YOLO algorithm focuses on detecting insulator defects within the intricate background of power transmission lines. To enhance its functionality, we propose three improved modules. Firstly, the Insulator Lack-Global Attention Mechanism (IL-GAM) addresses issues such as the mutual influence of weights and loss of detailed information in the original module. Secondly, the Insulator Lack-C3 (IL-C3) module is designed to emphasize key information while preserving feature extraction and fusion. Lastly, the Insulator Lack-SPPFCSPC (IL-SPPFCSPC) module enhances attention to both key and global information while extracting effective information from multi-scale features. Experimental results demonstrate that IL-YOLO achieves a detection accuracy of 91.2%, marking a 3.6% improvement compared to the YOLOv5 algorithm. Furthermore, precision improves by 0.5%, recall increases by 6.3%, and the F1 score sees a boost of 3.8%. Notably, IL-YOLO achieves a frame rate of 90 frames per second (FPS), showcasing its capacity for real-time detection. Additional experiments affirm IL-YOLO’s accuracy in completing insulator defect detection tasks in both general and complex backgrounds, highlighting its substantial advantages in addressing complex background and multi-target challenges.

This paper provides a study of the latest target (object) detection algorithms for underwater wireless sensor networks (UWSNs). To ensure selection of the latest and state-of-the-art algorithms, only algorithms developed in the last seven years are taken into account that are not entirely addressed by the existing surveys. These algorithms are classified based on their architecture and methodologies of operation and their applications are described that are helpful in their selection in a diverse set of applications. The merits and demerits of the algorithms are also addressed that are helpful to improve their performance in future investigation. Moreover, a comparative analysis of the described algorithms is also given that further provides an insight to their selection in various applications and future enhancement. A depiction of the addressed algorithms in various applications based on publication count over the latest decade (2023-2013) is also given using the IEEE database that is helpful in knowing their future application trend. Finally, the challenges associated with the underwater target detection are highlighted and the future research paradigms are identified. The conducted study is helpful in providing a thorough analysis of the underwater target detection algorithms, their feasibility in various applications with future challenges and defined strategies for further investigation.

Periodontitis is a high prevalence dental disease caused by bacterial infection of the bone that surrounds the tooth. Early detection and precision treatment can prevent more severe symptoms such as tooth loss. Traditionally, periodontal disease is identified and labeled manually by dental professionals. The task requires expertise and extensive experience, and it is highly repetitive and time-consuming. The aim of this study is to explore the application of AI in the field of dental medicine. With the inherent learning capabilities, AI exhibits remarkable proficiency in processing extensive datasets and effectively managing repetitive tasks. This is particularly advantageous in professions demanding extensive experiential knowledge, such as dentistry. By harnessing AI, the potential arises to amplify process efficiency and velocity. In this study, bitewing radiographs are used as the image source, and there are two major steps to detect the dental symptoms including 1) tooth position identification; and 2) symptom identification. The study combines image enhancement techniques and tooth position identification using Gaussian filtering and adaptive binarization for data preprocessing, facilitated by the YOLOv4 model to precisely mark tooth positions. The subsequent step enhances symptom area visibility via contrast enhancement, utilizing a CNN model, particularly the AlexNet model, with significant improvements in caries recognition accuracy (92.85%) and restorations recognition accuracy (96.55%) compared to prior research. Moreover, the inclusion of periodontal disease symptoms achieves an accuracy of 91.13%. By harnessing deep learning techniques based on CNN models, this research enhances diagnostic precision, reduces errors, and increases efficiency for dentists, thereby providing meticulous and swift patient care. This innovation not only saves time but also has the potential for widespread implementation in remote and preventive medicine, aligning with the aspiration of universal health care accessibility.

In this study, we propose a novel lightweight detection model for rebar counting, which is rectified mobilenet lightweight feature pyramid network based on YOLO (RM-LFPN-YOLO). The model incorporates a lightweight backbone network that integrates the coordinate attention (CA) mechanism, a lightweight feature pyramid network (LFPN), and a loss function that combines focal loss and efficient intersection over union (EIOU) loss, all meticulously designed to enhance the model’s performance. Experimental results demonstrate that our improved algorithm, with a mere 25.08M parameters, computes efficiently at 7.60G with an input size of 416 pixels. Additionally, it achieves an impressive average precision (AP) of 99.03% at an IOU of 0.5. The proposed lightweight model can be deployed on embedded devices and achieve efficient rebar detection and counting performance.

Object detection is a widely applied approach in addressing many real-world computer vision challenges. Despite its importance, object detection is computationally intensive and time-consuming, even with advanced CPU-GPU combinations. With the rise of edge computing and smaller AI accelerators, there is an increasing need to deploy efficient object detection applications on near-edge devices, such as drones and autonomous vehicles. However, these applications often face significant challenges and performance limitations due to restricted computational resources. Traditional object detection methods, e.g., Regional Convolutional Neural Network (RCNN) and You Only Look Once (YOLO), have extensive weight parameters, leading to high demands on memory and computing resources. Therefore, it is important to compress and optimize object detection models by reducing both the size and the number of weight parameters. This review article delves into the current state of object detection methods and simplification strategies, with a focus on deep-learning compression techniques. We investigate various approaches to mitigate these computational challenges, including replacing the regional proposal network (RPN), compressing model backbones, and modifying model heads, specifically for near-edge devices with limited and energy-efficient CPUs and GPUs. While simplifying object detection models is expected to reduce processing time significantly, it can also negatively impact model accuracy. Therefore, we discuss the ongoing challenge of finding the optimal model compression that balances speed while maintaining high accuracy.

The objects in optical high-resolution remote sensing images (HRRSIs) are usually tiny, dense, and exist in complex backgrounds, which brings great challenges to accurate object detection. This article presents an improved YOLOv5s network-based technique for remote sensing object recognition to overcome these difficulties. First, unnecessary residual modules are pruned from the cross-stage partial layer of conventional YOLOv5s and a refined residual coordinate attention module is incorporated to enhance the feature representation of the densely packed small objects in HRRSIs by introducing the residual structure and the mix pooling operation instead of the existing average pooling. Second, since various scales of objects are present in HRRSIs, the algorithm of differential evolution is adopted to replace the traditional K-means for generating a variety of anchor boxes in different sizes. Third, we replace the commonly used complete intersection over union (IoU) loss function in YOLOv5s with the AW-IoU loss function based on both α-IoU and wise-IoU. AW-IoU could expedite bounding box regression and focus more on regular anchor boxes. Finally, instead of nonmaximum suppression (NMS), the SCYLLA (S-IoU) soft-NMS is employed to eliminate the redundant duplicate boxes to detect the dense objects in remote sensing images. Experimental results on the NWPU VHR-10 dataset demonstrate that the proposed YOLOv5s method performs well compared with state-of-the-art algorithms. [ABSTRACT FROM AUTHOR]

Inspired by the recent progress of multimodal fusion in a variety of computer vision tasks, this letter aims to propose a two-stream fusion crater detection network (TFCDNet). Toward this end, near-infrared (IR) images and digital elevation maps (DEMs) in the feature domain are appropriately fused to boost the performance of crater detection (CD). The proposed TFCDNet includes a powerful feature-coding module that can effectively extract and fuse multimodal features. The comprehensively conducted experiments on both optical-DEM paired lunar crater detection dataset (ODPLCD) and Mars day CD (MDCD) datasets reveal that the proposed TFCDNet is capable of being more competitive than the state of the arts. As a result, this work is anticipated to spark some new thinking in CD. Relevant data in this letter can be downloaded from the website https://doi.org/10.57760/sciencedb.o00009.00312. [ABSTRACT FROM AUTHOR]

Intracranial hematoma due to traumatic brain injury is a serious health concern with rates of morbidity and mortality that are increasing worldwide. Manual identification is slow, subject to observer variabilities, and the existing automated techniques for intracranial hematoma detection in non-contrast computed tomography images cannot effectively detect multiple lesions of irregular sizes and shapes. Therefore, a computer-aided system using different window settings, YOLOv5s, cascaded attention module, and spatial pyramid pooling-fast is proposed to detect hematoma types, namely acute intraparenchymal, intraventricular, subdural, epidural, subarachnoid, and chronic subdural. Firstly, the computed tomography images are pre-processed using a window-based stacking approach wherein a three-channel image is generated by stacking grayscale images obtained with the aid of multiple window settings, i.e, brain, bone, and subdural. Secondly, a cascaded attention module is constructed in the neck of the YOLOv5s model to improve its detection performance by placing the convolution block attention module in serial with the efficient channel attention module. The cascaded attention module enriches the feature representation of various hematoma types in complex backgrounds especially when they are small and inconspicuous. The spatial pyramid pooling is replaced by a spatial pyramid pooling-fast to reduce the computational parameters and accelerate the feature fusion ability. The proposed deep learning model is trained, validated, and tested with 15,921 images from the brain haemorrhage extended dataset and it achieved overall precision, recall, F1-score, and mean average precision at 0.5, and mean average precision at 0.5:0.95 of 0.935, 0.908, 0.921, 0.943 and 0.65 respectively. The experimental results show that in comparison to the original YOLOv5s model and state-of-the-art methods, the model was able to localize and classify the acute or chronic instances of five hematoma subtypes in an individual image with improved precision and recall values. Hence the proposed system can be used in hospitals for the early and accurate detection of hematoma.

Research has shown that cognitive engagement plays a key role in effective learning, resulting in extensive efforts have been devoted to measuring it. Whereas most of the literature explores manual methods to measure cognitive engagement, the research on automatic detection of cognitive engagement levels in real classrooms is very limited. Automatic detection of cognitive engagement has been a problem for a long time due to the lack of behavior annotation guidance and effective detection algorithms. For the first challenge, a theory of cognitive engagement called Interactive-Constructive-Active-Passive-Disengage (ICAPD) is proposed in this paper. ICAPD links visual behaviors with cognitive engagement in the classroom. According to the ICAPD framework, a cognitive engagement dataset is constructed to train the detection model. To tackle the second challenge, the simAM-based You Only Look Once version 8 Nano (simAM-YOLOv8n) model is designed, simAM-YOLOv8n utilizes the simAM attention module to strengthen feature extraction and detect different levels of cognitive engagement precisely and efficiently. The experimental results on the self-build dataset have demonstrated the effectiveness of the proposed theory framework and detection algorithm, indicating that the proposed methodology could be used to detect real-time cognitive engagement in the classroom scenario. This work has the potential to help teachers to carry out learning analysis and instructional adjustments.

In recent years, automotive radar has attracted considerable attention due to the growing interest in autonomous driving technologies. Acquiring situational awareness using multimodal data collected at high sampling rates by various sensing devices including cameras, LiDAR, and radar requires considerable power, memory and compute resources which are often limited at an edge device. In this paper, we present a novel adaptive radar sub-sampling algorithm designed to identify regions that require more detailed/accurate reconstruction based on the information about prior environmental conditions, enabling near-optimal performance at considerably lower effective sampling rates. Designed to robustly perform under variable weather conditions, the algorithm was shown on the Oxford radar dataset to achieve accurate scene reconstruction utilizing only 10% of the collected samples in good weather. In the case of the RADIATE dataset acquired during extreme weather conditions (snow, fog), only 20% of the samples were sufficient to enable robust scene reconstruction. A further modification of the algorithm incorporates object motion to enable reliable identification of regions that require attention. This includes monitoring possible future occlusions caused by the objects detected in the present frame. Finally, we train a YOLO network on the RADIATE dataset to perform object detection, obtaining 6.6% AP50 improvement over the baseline Faster R-CNN network.

The surface cracks on high-speed railway ballastless track slabs directly influence their lifespan, while the efficiency of damage detection and maintenance is crucial for ensuring operational safety. Leveraging deep learning image processing technology can significantly enhance detection efficiency. Therefore, in response to the specific attributes of ballastless track slab crack detection, this paper introduces the RSG-YOLO model. By implementing a reparameterized dual-fused feature pyramid structure, we bolster the network’s feature extraction capacity and curtail the loss of crack features during extraction. SIoU is used to replace CIoU to optimize the bounding box regression loss function, reduce the degrees of freedom of the loss function, and improve the convergence speed The GAM attention mechanism is integrated to heighten the model’s responsiveness to diverse channel information. The proposed RSG-YOLO model was evaluated against mainstream models in the field of crack detection. The results demonstrated improved detection accuracy and recall rates. Specifically, when compared to baseline models, our approach exhibited significant advancements in reducing both missed detections and false alarms. These improvements were quantified by a 4.34% increase in crack detection accuracy and a 3.08% rise in mAP_0.5. Consequently, the RSG-YOLO model effectively enables precise detection of track slab cracks.