YOLO-ACN: Focusing on Small Target and Occluded Object Detection

Small target detection is a challenging task in very-high-resolution (VHR) optical remote sensing imagery, because small targets occupy a minuscule number of pixels and are easily disturbed by backgrounds or occluded by others. Although current convolutional neural network (CNN)-based approaches perform well when detecting normal objects, they are barely suitable for detecting small ones. Two practical problems stand in their way. First, current CNN-based approaches are not specifically designed for the minuscule size of small targets (~15 or ~10 pixels in extent). Second, no well-established data sets include labeled small targets and establishing one from scratch is labor-intensive and time-consuming. To address these two issues, we propose an approach that combines Sig-NMS-based Faster R-CNN with transfer learning. Sig-NMS replaces traditional non-maximum suppression (NMS) in the stage of region proposal network and decreases the possibility of missing small targets. Transfer learning can effectively label remote sensing images by automatically annotating both object classes and object locations. We conduct an experiment on three data sets of VHR optical remote sensing images, RSOD, LEVIR, and NWPU VHR-10, to validate our approach. The results demonstrate that the proposed approach can effectively detect small targets in the VHR optical remote sensing images of about $10\times 10$ pixels and automatically label small targets as well. In addition, our method presents better mean average precisions than other state-of-the-art methods: 1.5% higher when performing on the RSOD data set, 17.8% higher on the LEVIR data set, and 3.8% higher on NWPU VHR-10.

Traditional cloud computing faces issues like high bandwidth pressure, service delay, and energy consumption, while simple edge computing struggles with complex tasks due to limited resources. Thus, a collaborative edge cloud data processing and adaptive optimization method for intelligent perception systems is proposed. A three-layer model (terminal, edge cluster, cloud center) formalizes tasks, communication, and computing resources for dynamic optimization. A collaborative data processing flow is designed, with lightweight terminal preprocessing based on priority and an intelligent edge caching mechanism to reduce redundant data transmission. Task offloading and resource allocation are framed as reinforcement learning tasks, using the PPO algorithm for intelligent decision-making. Experiments show average processing latencies of 150.2-260.1 ms under varying loads, achieving high energy efficiency (98.5 FLOPs/J, Floating-point Operations per Joule) and object detection accuracy (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$97.2 \% \text{mAP}$</tex>, mean Average Precision). The method outperforms others in latency and balances energy efficiency and accuracy, efficiently utilizing resources amid dynamic tasks and uncertainties.

Nowadays, increasing efficiency in the service sector by objectively monitoring and tracking business workflows has become a critical requirement for the sustainability of businesses. Traditional monitoring methods are insufficient for providing sustainable performance analysis due to their time-consuming nature and reliance on subjective human judgment. The aim of this study is to develop a system that automatically detects and classifies employee behavior using computer vision and deep learning techniques. As part of the study, data was collected using a camera placed in a real bakery environment. Five basic classes were labeled on the created data set: cleaning, product interaction, computer use, phone use, and money interaction. The current YOLOv11 (You Only Look Once) architecture, which offers high speed and accuracy for object detection and classification, was used. According to the experimental results obtained from training the model, the system demonstrated high performance, achieving 0.9552 Precision, 0.9324 Recall, 0.9437 F1-Score, and 0.9644 mAP@50 values. These results demonstrate that the proposed system can detect employee behaviors in real-time with a high accuracy rate, allowing it to be used as an effective tool in workplace productivity enhancement and performance evaluation processes.

The ability to detect small targets and the speed of the target detector are very important for the application of remote sensing image detection, and in this paper, we propose an effective and efficient method (named CISPNet) with high detection accuracy and compact architecture. In particular, according to the characteristics of the data, we apply a context information scene perception (CISP) module to obtain the contextual information for targets of different scales and use k-means clustering to set the aspect ratios and size of the default boxes. The proposed method inherits the network structure of Single Shot MultiBox Detector (SSD) and introduces the CISP module into it. We create a dataset in the Pascal Visual Object Classes (VOC) format, annotated with the three types of detection targets, aircraft, ship, and oiltanker. Experimental results on our remote sensing image dataset as well as the Northwestern Polytechnical University very-high-resolution (NWPU VRH-10) dataset demonstrate that the proposed CISPNet performs much better than the original SSD and other detectors especially for small objects. Specifically, our network can achieve 80.34% mean average precision (mAP) at the speed of 50.7 frames per second (FPS) with the input size 300 × 300 pixels on the remote sensing image dataset. On extended experiments, the performance of CISPNet in fuzzy target detection in remote sensing image is better than that of SSD.

With the rapid development of optical remote sensing, it is urgent to find a reliable target detection method. Compared with traditional detection algorithms, a convolutional neural network has attracted considerable attention owing to its efficiency and high transitivity. However, different from general images, remote sensing images contain complex background information and dense small targets with changeable directions that make detection very challenging. To solve these problems and provide a stable and high-performance detection method, a rotated saliency fusion object detection (RSD) model based on “you only look once” (YOLO)v4 is established. First, salient image fusion technology is used to magnify target information. Second, the angle variable and rotated non-maximal suppression is introduced to improve the accuracy of rotated object detection by including the detection of dense objects. Third, the network structure is enhanced to improve the performance of small-target detection. Finally, the k-means algorithm and data enhancement are introduced to increase the robustness of the model. Extensive experiments demonstrate the superiority of the proposed model in detection speed and accuracy. The mean average precision of the proposed RSD model reaches 97.32% for the remote sensing images in a harbor area with an average detection speed of 13.41 s − 1.

Object detection for mobile devices is meaningful especially in the field of IoT. Limited by computing power and network transmission, it's challenging to get high accuracy in mobile object detection. To solve this question, this article designs a system that enables high accuracy object detection running at 30fps for 720p videos. The system employs the object tracking technique, uses the caching technique, decouples the rendering pipeline from the offloading pipeline, and uses dynamic RoI encoding technique to get high detection accuracy. The result of the experiment shows that it can get 88% detection success rate. And it can also increase the detection accuracy by 17.7% and decrease the bandwidth by 52.6%.

In response to the low inefficiency and high misjudgement rate of manually observing grape leaf diseases, an improved YOLOv8n grape leaf disease detection model NMA-YOLOv8n is proposed. Firstly, the global nonlinear attention NBL was introduced in the neck network, which enhances the backbone feature extraction capability by fusing the local and non-local attention, enabling the network to equally focus on small and normal targets. Secondly, the MPDCIoU loss function is designed to replace the original Bbox loss function, improving the regression accuracy of the bounding box. At the output end, the detection performance of the algorithm for small targets is improved by designing the AFPN small target detection head. The experimental results show that the NMA-YOLOv8n model mAP@0.5 reaches 94.7%, 1.7% higher than YOLOv8n; the FPS reaches 124.6 frames/sec, which can meet the real-time detection requirements, and has higher detection accuracy and speed compared with other five mainstream target detection models.NMA-YOLOv8n provides grape disease detection with a better method, which has certain significance for the prevention and control of grape diseases.

Background and Objective: Ensuring safety in high-risk environments like industrial zones and construction zones is a basic requirement. Here, the use of helmets is a mandatory protocol to protect the workers from head injuries. But even if strict rules have been complied with in these industrial areas, real-world compliance is inconsistent due to negligence and ignorance. Traditional methods of surveillance-based monitoring are both resource-intensive as well as can be easily prone to human oversight. This research proposes a transformer-based deep learning solution using DETR (DEtection TRansformer) for automated helmet detection. This aims to modernise and automate safety compliance monitoring systems. Methods: The proposed system leverages DETR, a novel object detection architecture that combines convolutional neural networks with Vision Transformers to eliminate the need for region proposal networks and non-maximum suppression. We calibrate DETR on a publicly available Kaggle dataset consisting of 764 images annotated in PASCAL VOC format across two classes: With Helmet and Without Helmet. The model was trained using the PyTorch framework, with strategic data augmentation applied to enhance generalisation. Performance was evaluated using mean Average Precision (mAP), precision, recall, and inference time. Results: The DETR-based model achieved promising results in distinguishing helmet usage, delivering high accuracy and reliable localisation even with a relatively small dataset. The proposed attention-based model achieved competitive results in hard cases, where it performed better than the traditional CNN-based approaches. Conclusion: The study demonstrates that transformer-based object detection systems are successful in meeting real-world safety compliance tasks. The DETR architecture provides a scalable intelligence-based system for real-time helmet monitoring and can be used in automated surveillance systems in various industrial scenarios.

To improve the detection accuracy and efficiency of small targets, overlapping targets, and complex backgrounds in blood cell images, this paper proposes an improved model based on the YOLOv12n lightweight architecture. The model integrates three core modules: the A2C2f_DYT module introduces a dynamic nonlinear mechanism and multi-head attention to enhance feature representation; the C2BRA module adopts dual-layer routing attention to achieve multi-scale semantic fusion; and the DySample module enhances the restoration of fine-grained spatial information through dynamic point sampling. Comparative experiments conducted on the standard blood cell dataset show that the improved model achieves high detection accuracy (mAP@0.5 of 97.1%) while maintaining low computational complexity (6.0 GFLOPS) and a lightweight parameter size (2.79M), outperforming mainstream models such as YOLOv8n, YOLOv10n, and YOLOv12n. Furthermore, the model demonstrates excellent accuracy and recall rates in multi-class detection, including WBCs, RBCs, and platelets, showcasing superior performance in small target detection and class adaptability. Ablation experiments further validate the positive contribution of each module to detection performance. This study presents a blood cell detection solution with high accuracy, a lightweight design, and strong robustness, making it suitable for deployment on edge devices and in intelligent medical scenarios.

Cotton, as a globally important economic crop, requires early and accurate disease detection to ensure stable yield and promote sustainable development. However, due to the small size of certain leaf lesions, traditional detection methods often suffer from missed or false detections. To address this issue, we propose an improved YOLOv8-based model, CM-YOLO, aimed at enhancing the detection performance for small cotton leaf disease targets. Specifically, the SS2D module from VMamba is introduced into the backbone network to achieve comprehensive feature extraction through multi-directional scanning. Furthermore, the MSDA module is embedded prior to the SPPF module to reduce performance degradation caused by redundant computations and to enhance the model’s focus on critical small targets. Finally, the original bounding box loss function is replaced with DIoU, enabling precise localization of small targets by optimizing anchor center point distances and accelerating model convergence. Experimental results demonstrate that CM-YOLO achieves superior performance in cotton leaf disease detection, with an mAP50 of 0.933 and a recall of 0.891. Compared with state-of-the-art methods, YOLOv8n and YOLOv11n achieve mAP50 values of 0.874 and 0.930, respectively, both lower than CM-YOLO, thereby validating the effectiveness of the proposed method. Additionally, generalization experiments indicate that the model maintains high detection accuracy and robustness across different plant datasets, highlighting its strong applicability in complex scenarios and providing a valuable reference for intelligent agricultural disease detection research.

Wind turbines, as the core components of wind power generation systems, play a crucial role in determining the overall generation efficiency and operational safety. However, the challenges posed by complex backgrounds, significant variations in the scale of wind turbine targets, and arbitrary orientations in unmanned aerial vehicle (UAV) remote sensing images have significantly increased the difficulty of real-time wind turbine detection. To address these challenges, this paper introduces an enhanced multi-view onshore wind turbine remote sensing detection algorithm for UAVs based on YOLOv7, termed EGRN-YOLO. Firstly, the lightweight network EfficientNetV2 is utilized as the feature extraction backbone to reduce the number of model parameters and computational load. Secondly, to further minimize the model parameters, the GhostSPPCSPC module is introduced to replace the original pyramid module SPPCSPC. To maintain a balance between accuracy and parameter efficiency, the CAM attention mechanism is integrated to create the GhostSPPCSPC_CAM module, which effectively expands the network’s receptive field and enhances its deep learning capabilities. Subsequently, to improve the network’s focus on small targets and reduce the redundancy of gradient information during neural network inference, the RepNCSPELAN4_KAN module is proposed and integrated into the neck network. Finally, the NWD loss function is introduced to replace the CIoU loss function, thereby improving the accuracy of positive and negative sample allocation for small targets. Experimental results on the self-constructed VOC_WIND_TURBINE dataset demonstrate that the improved EGRN-YOLO model, compared to the benchmark YOLOv7 model, achieves significant improvements of 1.7% in mean average precision (mAP), 3.3% in precision, 1.8% in recall, and 18.1% in inference time. These results indicate a well-balanced trade-off between high accuracy and model efficiency, effectively meeting the demands for identifying wind turbines in complex scenes within remote sensing images.

Wind energy is a renewable resource with abundant reserves, and its sustainable development and utilization are crucial. The components of wind turbines, particularly the blades and various surfaces, require meticulous defect detection and maintenance due to their significance. The operational status of wind turbine generators directly impacts the efficiency and safe operation of wind farms. Traditional surface defect detection methods for wind turbines often involve manual operations, which suffer from issues such as high subjectivity, elevated risks, low accuracy, and inefficiency. The emergence of computer vision technologies based on deep learning has provided a novel approach to surface defect detection in wind turbines. However, existing datasets designed for wind turbine surface defects exhibit overall category scarcity and an imbalance in samples between categories. The algorithms designed face challenges, with low detection rates for small samples. Hence, this study first constructs a benchmark dataset for wind turbine surface defects comprising seven categories that encompass all common surface defects. Simultaneously, a wind turbine surface defect detection algorithm based on improved YOLOv5 is designed. Initially, a multi-scale copy-paste data augmentation method is proposed, introducing scale factors to randomly resize the bounding boxes before copy-pasting. This alleviates sample imbalances and significantly enhances the algorithm’s detection capabilities for targets of different sizes. Subsequently, a dynamic label assignment strategy based on the Hungarian algorithm is introduced that calculates the matching costs by weighing different losses, enhancing the network’s ability to learn positive and negative samples. To address overfitting and misrecognition resulting from strong data augmentation, a two-stage progressive training method is proposed, aiding the model’s natural convergence and improving generalization performance. Furthermore, a multi-scenario negative-sample-guided learning method is introduced that involves incorporating unlabeled background images from various scenarios into training, guiding the model to learn negative samples and reducing misrecognition. Finally, slicing-aided hyper inference is introduced, facilitating large-scale inference for wind turbine surface defects in actual industrial scenarios. The improved algorithm demonstrates a 3.1% increase in the mean average precision (mAP) on the custom dataset, achieving 95.7% accuracy in mAP_50 (the IoU threshold is half of the mAP). Notably, the mAPs for small, medium, and large targets increase by 18.6%, 16.4%, and 6.8%, respectively. The experimental results indicate that the enhanced algorithm exhibits high detection accuracy, providing a new and more efficient solution for the field of wind turbine surface defect detection.

Object detection and recognition is an essential task in computer vision with numerous real-world applications such as surveillance, self-driving cars, and robotics. In recent years, deep learning-based approaches have significantly improved the accuracy and speed of object detection and recognition. The You Only Look Once version 3 (YOLOv3) algorithm is a popular deep learning-based approach that can detect and recognize objects in real-time. The Common Objects in Context (COCO) dataset is a large-scale dataset with over 330,000 labeled images and more than 2.5 million object instances, making it a popular choice for object detection and recognition tasks. In this paper, we propose a deep learning-based approach for real-time object detection and recognition using the YOLOv3 architecture and COCO dataset. We evaluate our approach based on several performance metrics, including mean average precision (mAP), frames per second (FPS), total object detection time, object detection accuracy, false positive rate, number of detected objects, and mean intersection over union (mIoU). Our results show that our approach achieves a mean average precision of 0.76 on the COCO dataset and a real-time performance of 40 frames per second on a single GPU. Additionally, our approach achieves an object detection accuracy of 93.5%, a false positive rate of 6.5%, and a mean intersection over union of 0.65. Our proposed approach shows promising results for real-time object detection and recognition and can be applied to various real-world applications.

Recently, drones have been applied and used in various tasks in many fields. Due to the development of camera technology and hardware, it is possible to detect objects using deep learning technology in real time. Most of the object sizes of videos and images taken by drones are very small objects. The dataset that collects these image data is the VisDrone dataset. As deep learning technology develops a lot, object detection accuracy is getting higher and higher. However, it is still hard to perform object detection in real-time and show high object detection accuracy. In this paper, we combine YOLOv5 and CBAM to increase object detection accuracy by focusing more on the features required for object detection. The model is trained using the VisDrone dataset, and the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{mAP}$</tex> value is measured at 22. <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$56\text{mAP}$</tex> ,which is 2. <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$45\text{mAP}$</tex> higher than the original YOLOv5.

Forest fires have continually endangered personal safety and social property. To reduce the occurrences of forest fires, it is essential to detect forest fire smoke accurately and quickly. Traditional forest fire smoke detection based on convolutional neural networks (CNNs) needs many hand-designed components and shows poor ability to detect small and inconspicuous smoke in complex forest scenes. Therefore, we propose an improved early forest fire smoke detection model based on deformable transformer for end-to-end object detection (deformable DETR). We use deformable DETR as a baseline containing the best sparse spatial sampling for smoke with deformable convolution and relation modeling capability of the transformer. We integrate a Multi-scale Context Contrasted Local Feature module (MCCL) and a Dense Pyramid Pooling module (DPPM) into the feature extraction module for perceiving features of small or inconspicuous smoke. To improve detection accuracy and reduce false and missed detections, we propose an iterative bounding box combination method to generate precise bounding boxes which can cover the entire smoke object. In addition, we evaluate the proposed approach using a quantitative and qualitative self-made forest fire smoke dataset, which includes forest fire smoke images of different scales. Extensive experiments show that our improved model’s forest fire smoke detection accuracy is significantly higher than that of the mainstream models. Compared with deformable DETR, our model shows better performance with improvement of mAP (mean average precision) by 4.2%, APS (AP for small objects) by 5.1%, and other metrics by 2% to 3%. Our model is adequate for early forest fire smoke detection with high detection accuracy of different-scale smoke objects.