Search

Due to the relatively high shooting altitude of unmanned aerial vehicles (UAV), the captured images often contain a multitude of small‐scale targets. To solve the problems of small target scale, lack of semantic information, and high miss detection in drone target detection, in this paper we proposed a more effective unmanned aerial vehicle small target detection algorithm(MEU‐YOLOv5) based on YOLOv5s. Firstly, an efficient global contextual module is proposed to enhance the algorithm's performance in feature extraction while reducing the excessive loss of shallow features. Secondly, a small‐scale target detector is added to enhance the algorithm's detection capability for smaller targets. Lastly, a recursive multi‐level feature fusion path is introduced to better fuse the shallow and deep features of the images, reducing overfitting and improving the algorithm's generalizability and robustness. Experimental results demonstrated that compared to YOLOv5s, MEU‐YOLOv5 achieves a 7.4% improvement in mAP@0.5 and a 4.9% improvement in mAP@0.5:0.95. Additionally, the overall performance of this algorithm surpassed various algorithms in the YOLO series, including YOLOv3, YOLOv5l, YOLOv5m, and YOLOv8s. © 2024 Institute of Electrical Engineers of Japan. Published by Wiley Periodicals LLC. [ABSTRACT FROM AUTHOR]

Landslides are among the most dangerous and catastrophic natural hazards with countless concerns. In disaster rescue operations, fast and precise identification of landslides is necessary for timely and effective preventive actions. The landslide risk is anticipated to be reduced through their prediction, monitoring, and accurate detection using remote sensing technology. Moreover, deep learning algorithms have shown excellent improvement in various remote sensing applications. Recent scientific and intelligent technological innovations are needed to be applied to disaster management and assessment, particularly landslides. Therefore, this study aims to extract the landslide hazard information from multiple data sources, i.e., satellite and unmanned aerial vehicle (UAV) images, using a single staged object detection model, i.e., YOLOv5, YOLOv6, YOLOv7, and YOLOv8. The data from distinct platforms are utilized to infer the synergies between them. The results of each database are evaluated quantitatively using standard methods, i.e., precision, recall, f-score, and mean average precision, whereas visual analysis of results is conducted for qualitative assessment. Based on the experimental results, the highest f-score is represented by YOLOv7 (0.995) and YOLOv5 (0.921) for satellite and UAV-based data, respectively. The quantitative results are further compared with previous research work to exhibit the novelty and competence of the proposed research. Our work demonstrates the application and feasibility of the YOLO model in landslide information extraction for quick hazard recovery operations. [ABSTRACT FROM AUTHOR]

Steel surface defect detection poses a significant challenge in the steel industry, aiming to enhance product quality and production efficiency. Traditional mechanical and optical detection methods exhibit relatively low efficiency and poor real-time performance in detecting acceptable defects on the surface of steel strips. This paper proposes a new model named Hyper-YOLO for steel surface defect detection in steel strips. Firstly, the CSP module in the conventional YOLO backbone network is replaced with the Ghost module. The Ghost module, a lightweight convolutional module, enhances model efficiency by reducing parameter count and computational load while maintaining satisfactory performance. Secondly, researchers replace the PAFPN module in YOLO V5 in the bottleneck section with the Hyper FPN module. Hyper FPN, an improved feature pyramid network module, leverages features at different scales for multi-level feature fusion, enhancing the model's capability to detect targets at various scales. Lastly, improvements are made in the loss functions for both training and prediction stages. The α-CIoU loss function is introduced during training to substitute the original CIoU loss function, and the α-DIoU loss function is utilized during prediction instead of the original DIoU loss function. These enhanced loss functions effectively measure the accuracy and position precision of target boxes, thereby improving the detection performance of the model. Through these enhancements, the Hyper YOLO model achieves an overall performance improvement of 4.58% over the baseline model. This indicates that Hyper YOLO performs outstanding surface defect detection in steel strips, providing innovative insights for the YOLO V5 model. These improvements not only elevate the accuracy and efficiency of the model but also hold significant guidance for similar research and applications. [ABSTRACT FROM AUTHOR]

This review presents an investigation into the incremental advancements in the YOLO (You Only Look Once) architecture and its derivatives, with a specific focus on their pivotal contributions to improving quality inspection within the photovoltaic (PV) domain. YOLO's single-stage approach to object detection has made it a preferred option due to its efficiency. The review unearths key drivers of success in each variant, from path aggregation networks to generalised efficient layer aggregation architectures and programmable gradient information, presented in the latest variant, YOLOv10, released in May 2024. Looking ahead, the review predicts a significant trend in future research, indicating a shift toward refining YOLO variants to tackle a wider array of PV fault scenarios. While current discussions mainly centre on micro-crack detection, there is an acknowledged opportunity for expansion. Researchers are expected to delve deeper into attention mechanisms within the YOLO architecture, recognising their potential to greatly enhance detection capabilities, particularly for subtle and intricate faults. [ABSTRACT FROM AUTHOR]

Engineering vehicles play a vital role in supporting construction projects. However, due to their substantial size, heavy tonnage, and significant blind spots while in motion, they present a potential threat to road maintenance, pedestrian safety, and the well-being of other vehicles. Hence, monitoring engineering vehicles holds considerable importance. This paper introduces an engineering vehicle detection model based on improved YOLOv6. First, a Swin Transformer is employed for feature extraction, capturing comprehensive image features to improve the detection capability of incomplete objects. Subsequently, the SimMIM self-supervised training paradigm is implemented to address challenges related to insufficient data and high labeling costs. Experimental results demonstrate the model's superior performance, with a m A P 50 : 95 value of 88.5% and m A P 50 value of 95.9% on the dataset of four types of engineering vehicles, surpassing existing mainstream models and proving its effectiveness in engineering vehicle detection. [ABSTRACT FROM AUTHOR]

Object monitoring in video analysis is a crucial component of smart surveillance systems, enabling the observation and analysis of customer behavior and interactions with various items. This research aims to present a vision-based approach to object monitoring specifically tailored for surveillance systems in smart shops, utilizing an intelligent Internet of Things device. The focus of this study revolves around monitoring jewelry objects within jewelry shops. To accomplish the object monitoring task, an object detector based on Yolo architecture is employed, allowing for the detection and localization of objects. The Yolo network is trained and validated using a customized dataset that encompasses two distinct classes of jewelry objects: rings and earrings. The experimental findings show that the suggested object monitoring approach fulfills the real-time requirements for monitoring objects within smart shops' surveillance systems. [ABSTRACT FROM AUTHOR]

Copyright of Przegląd Elektrotechniczny is the property of Przeglad Elektrotechniczny and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Computer Measurement & Control is the property of Magazine Agency of Computer Measurement & Control and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

The You Only Look Once (YOLO) object detection network has garnered widespread adoption in various industries, owing to its superior inference speed and robust detection capabilities. This model has proven invaluable in automating production processes such as material processing, machining, and quality inspection. However, as market competition intensifies, there is a constant demand for higher detection speed and accuracy. Current FPGA accelerators based on 8-bit quantization have struggled to meet these increasingly stringent performance requirements. In response, we present a novel 4-bit quantization-based neural network accelerator for the YOLOv5 model, designed to enhance real-time processing capabilities while maintaining high detection accuracy. To achieve effective model compression, we introduce an optimized quantization scheme that reduces the bit-width of the entire YOLO network—including the first layer—to 4 bits, with only a 1.5% degradation in mean Average Precision (mAP). For the hardware implementation, we propose a unified Digital Signal Processor (DSP) packing scheme, coupled with a novel parity adder tree architecture that accommodates the proposed quantization strategies. This approach efficiently reduces on-chip DSP utilization by 50%, offering a significant improvement in performance and resource efficiency. Experimental results show that the industrial object detection system based on the proposed FPGA accelerator achieves a throughput of 808.6 GOPS and an efficiency of 0.49 GOPS/DSP for YOLOv5s on the ZCU102 board, which is 29% higher than a commercial FPGA accelerator design (Xilinx's Vitis AI). [ABSTRACT FROM AUTHOR]

To enhance user experiences of reaching destinations in large, complex buildings, we have developed a indoor navigation system using Unity and a smartphone called INSUS. It can reset the user location using a quick response (QR) code to reduce the loss of direction of the user during navigation. However, this approach needs a number of QR code sheets to be prepared in the field, causing extra loads at implementation. In this paper, we propose another reset method to reduce loads by recognizing information of naturally installed signs in the field using object detection and Optical Character Recognition (OCR) technologies. A lot of signs exist in a building, containing texts such as room numbers, room names, and floor numbers. In the proposal, the Sign Image is taken with a smartphone, the sign is detected by YOLOv8, the text inside the sign is recognized by PaddleOCR, and it is compared with each record in the Room Database using Levenshtein distance. For evaluations, we applied the proposal in two buildings in Okayama University, Japan. The results show that YOLOv8 achieved mAP@0.5 0.995 and mAP@0.5:0.95 0.978 , and PaddleOCR could extract text in the sign image accurately with an averaged CER% lower than 10%. The combination of both YOLOv8 and PaddleOCR decreases the execution time by 6.71 s compared to the previous method. The results confirmed the effectiveness of the proposal. [ABSTRACT FROM AUTHOR]

Printed circuit boards (PCBs) provide stable connections between electronic components. However, defective printed circuit boards may cause the entire equipment system to malfunction, resulting in incalculable losses. Therefore, it is crucial to detect defective printed circuit boards during the generation process. Traditional detection methods have low accuracy in detecting subtle defects in complex background environments. In order to improve the detection accuracy of surface defects on industrial printed circuit boards, this paper proposes a residual large kernel network based on YOLOv5 (You Only Look Once version 5) for PCBs surface defect detection, called YOLO-RLC (You Only Look Once-Residual Large Kernel). Build a deep large kernel backbone to expand the effective field of view, capture global information more efficiently, and use 1 1 convolutions to balance the depth of the model, improving feature extraction efficiency through reparameterization methods. The neck network introduces a bidirectional weighted feature fusion network, combined with a brand-new noise filter and feature enhancement extractor, to eliminate noise information generated by information fusion and recalibrate information from different channels to improve the quality of deep features. Simplify the aspect ratio of the bounding box to alleviate the issue of specificity values. After training and testing on the PCB defect dataset, our method achieved an average accuracy of 97.3% (mAP50) after multiple experiments, which is 4.1% higher than YOLOv5-S, with an average accuracy of 97.6% and an Frames Per Second of 76.7. The comparative analysis also proves the superior performance and feasibility of YOLO-RLC in PCB defect detection. [ABSTRACT FROM AUTHOR]

Although pedestrian detection techniques are improving, this task is still challenging due to the problems of target occlusion, small targets, and complex pedestrian backgrounds in images of different scenes. As a result, the You Only Look Once algorithm exhibits lower detection accuracy. In this paper, a multisensor (MS) module that utilizes multiple dilated convolutions is proposed to sample feature images, avoiding information loss caused by repeated sampling, to improve the feature extraction and target detection performance of the algorithm. In addition, a lightweight shuffle-based efficient channel attention mechanism is introduced to conduct grouping in the channel dimension and perform parallel processing for each subfeature map channel. A new branch is introduced to enrich the channel feature information for multiscale feature representation. Finally, a distance intersection over union-based nonmaximum suppression (DIoU-NMS) method is introduced to minimize the occurrence of missed targets due to occlusion by taking the prediction box and ground truth box centroid locations information into account without increasing the computational cost over that of normal NMS. Our method is extensively evaluated on several challenging pedestrian detection datasets, achieving 87.73%, 34.7%, 93.96% and 95.23% mean average precision values on PASCAL VOC, MS COCO, Caltech Pedestrian and INRIA Person, which are respectively. The experimental results demonstrate the effectiveness of the method. [ABSTRACT FROM AUTHOR]

YOLO series algorithms are widely used in unmanned aerial vehicles (UAV) object detection scenarios due to their fast and lightweight properties. This article summarizes the key concepts in YOLO series algorithms, such as the anchor mechanism, feature fusion strategy, bounding box regression loss and so on and points out the advantages and improvement space of the YOLO series algorithms. Discussing the relevant technologies of the YOLOv1 to YOLOv7 series algorithms in detail in three parts: basic structure, strengths and weaknesses, and compares the algorithm performance. On this basis, combined with the challenges of object detection technology in UAV applications, various solutions for improving the YOLO series algorithms and applying them to UAV object detection scenarios are demonstrated. The improvement strategies, application scenarios, academic contributions and limitations of the algorithms are summarized. Finally, the future development directions and challenges of applying YOLO series algorithms to UAV object recognition are prospected. [ABSTRACT FROM AUTHOR]

The stator of a flat wire motor is the core component of new energy vehicles. However, detecting quality defects in the coating process in real-time is a challenge. Moreover, the number of defects is large, and the pixels of a single defect are very few, which make it difficult to distinguish the defect features and make accurate detection more difficult. To solve this problem, this article proposes the YOLOv8s-DFJA network. The network is based on YOLOv8s, which uses DSFI-HEAD to replace the original detection head, realizing task alignment. It enhances joint features between the classification task and localization task and improves the ability of network detection. The LEFG module replaces the C2f module in the backbone of the YOLOv8s network that reduces the redundant parameters brought by the traditional BottleNeck structure. It also enhances the feature extraction and gradient flow ability to achieve the lightweight of the network. For this research, we produced our own dataset of stator coating quality regarding flat wire motors. Data augmentation technology (Gaussian noise, adjusting brightness, etc.) enriches the dataset, to a certain extent, which improves the robustness and generalization ability of YOLOv8s-DFJA. The experimental results show that in the performance of YOLOv8s-DFJA compared with YOLOv8s, the mAP@.5 index increased by 6.4%, the precision index increased by 1.1%, the recall index increased by 8.1%, the FPS index increased by 9.8FPS/s, and the parameters decreased by 3 Mb. Therefore, YOLOv8s-DFJA can be better realize the fast and accurate detection of the stator coating quality of flat wire motors. [ABSTRACT FROM AUTHOR]

Purpose: Wood comprises different cell types, such as fibers, tracheids and vessels, defining its properties. Studying cells' shape, size, and arrangement in microscopy images is crucial for understanding wood characteristics. Typically, this involves macerating (soaking) samples in a solution to separate cells, then spreading them on slides for imaging with a microscope that covers a wide area, capturing thousands of cells. However, these cells often cluster and overlap in images, making the segmentation difficult and time-consuming using standard image-processing methods. Results: In this work, we developed an automatic deep learning segmentation approach that utilizes the one-stage YOLOv8 model for fast and accurate segmentation and characterization of macerated fiber and vessel form aspen trees in microscopy images. The model can analyze 32,640 x 25,920 pixels images and demonstrate effective cell detection and segmentation, achieving a mAP 0.5 - 0.95 of 78 %. To assess the model's robustness, we examined fibers from a genetically modified tree line known for longer fibers. The outcomes were comparable to previous manual measurements. Additionally, we created a user-friendly web application for image analysis and provided the code for use on Google Colab. Conclusion: By leveraging YOLOv8's advances, this work provides a deep learning solution to enable efficient quantification and analysis of wood cells suitable for practical applications. [ABSTRACT FROM AUTHOR]

Effective pest management in urban areas is critically challenged by the rapid proliferation of mosquito breeding sites. Traditional fumigation methods expose human operators to harmful chemicals, posing significant health risks ranging from respiratory problems to long-term chronic conditions. To address these issues, a novel fumigation robot equipped with sensor fusion technology for optimal pest control in urban landscapes is proposed. The proposed robot utilizes light detection and ranging data, depth camera inputs processed through the You Only Look Once version 8 (YOLOv8) algorithm for precise object recognition, and inertial measurement unit data. These technologies allow the robot to accurately identify and localize mosquito breeding hotspots using YOLOv8, achieving a precision of 0.81 and a mean average precision of 0.74. The integration of these advanced sensor technologies allows for detailed and reliable mapping, enhancing the robot's navigation through complex urban terrains and ensuring precise targeting of fumigation efforts. In a test case, the robot demonstrated a 62.5% increase in efficiency by significantly reducing chemical usage through targeted hotspot fumigation. By automating the detection and treatment of breeding sites, the proposed method boosts the efficiency and effectiveness of pest management operations and significantly diminishes the health risks associated with chemical exposure for human workers. This approach, featuring real-time object recognition and dynamic adaptation to environmental changes, represents a substantial advancement in urban pest management, offering a safer and more effective solution to a persistent public health issue. [ABSTRACT FROM AUTHOR]

The real-time detection of fruits and plants is a crucial aspect of digital agriculture, enhancing farming efficiency and productivity. This study addresses the challenge of embedding a real-time strawberry detection system in a small mobile robot operating within a greenhouse environment. The embedded system is based on the YOLO architecture running in a single GPU card, with the Open Neural Network Exchange (ONNX) representation being employed to accelerate the detection process. The experiments conducted in this study demonstrate that the proposed model achieves a mean average precision (mAP) of over 97%, processing eight frames per second for 512 × 512 pixel images. These results affirm the utility of the proposed approach in detecting strawberry plants in order to optimize the spraying process and avoid inflicting any harm on the plants. The goal of this research is to highlight the potential of integrating advanced detection algorithms into small-scale robotics, providing a viable solution for enhancing precision agriculture practices. [ABSTRACT FROM AUTHOR]

Wheat powdery mildew is a fungal disease that significantly impacts wheat yield and quality. Controlling this disease requires the use of resistant varieties, fungicides, crop rotation, and proper sanitation. Precision agriculture focuses on the strategic use of agricultural inputs to maximize benefits while minimizing environmental and human health effects. Object detection using computer vision enables selective spraying of pesticides, allowing for targeted application. Traditional detection methods rely on manually crafted features, while deep learning-based methods use deep neural networks to learn features autonomously from the data. You Look Only Once (YOLO) and other one-stage detectors are advantageous due to their speed and competition. This research aimed to design a model to detect powdery mildew in wheat using digital images. Multiple YOLOv8 models were trained with a custom dataset of images collected from trial areas at Tekirdag Namik Kemal University. The YOLOv8m model demonstrated the highest precision, recall, F1, and average precision values of 0.79, 0.74, 0.770, 0.76, and 0.35, respectively. [ABSTRACT FROM AUTHOR]

Augmented reality (AR) technology has opened up new possibilities for interactive and immersive learning experiences, where students can visualize abstract concepts and engage with digital objects in real-time. In this study, a real-time AR-based system is developed and, unlike other studies, incorporates sound effects that will play when the object is detected to support the teaching of English words to primary school students. Using the You Look Only Once (YOLO) v4 algorithm and the Common Objects in Context (COCO) dataset, the system was trained to detect objects with high accuracy. The proposed system was tested on 40 primary school students for 8 weeks and compared their learning outcomes with a control group using traditional teaching methods. According to the experimental results, students utilizing the AR-based system achieved an impressive 90% success rate in the end-of-term exam, whereas the control group attained a success rate of 65%. Our study demonstrates the potential of AR technology to enhance children's motivation and engagement in learning the English language. [ABSTRACT FROM AUTHOR]

The brain atlas, which provides information about the distribution of genes, proteins, neurons, or anatomical regions, plays a crucial role in contemporary neuroscience research. To analyze the spatial distribution of those substances based on images from different brain samples, we often need to warp and register individual brain images to a standard brain template. However, the process of warping and registration may lead to spatial errors, thereby severely reducing the accuracy of the analysis. To address this issue, we develop an automated method for segmenting neuropils in the Drosophila brain for fluorescence images from the FlyCircuit database. This technique allows future brain atlas studies to be conducted accurately at the individual level without warping and aligning to a standard brain template. Our method, LYNSU (Locating by YOLO and Segmenting by U-Net), consists of two stages. In the first stage, we use the YOLOv7 model to quickly locate neuropils and rapidly extract smallscale 3D images as input for the second stage model. This stage achieves a 99.4% accuracy rate in neuropil localization. In the second stage, we employ the 3D U-Net model to segment neuropils. LYNSU can achieve high accuracy in segmentation using a small training set consisting of images from merely 16 brains. We demonstrate LYNSU on six distinct neuropils or structures, achieving a high segmentation accuracy comparable to professional manual annotations with a 3D Intersection-over-Union (IoU) reaching up to 0.869. Our method takes only about 7 s to segment a neuropil while achieving a similar level of performance as the human annotators. To demonstrate a use case of LYNSU, we applied it to all female Drosophila brains from the FlyCircuit database to investigate the asymmetry of the mushroom bodies (MBs), the learning center of fruit flies. We used LYNSU to segment bilateral MBs and compare the volumes between left and right for each individual. Notably, of 8,703 valid brain samples, 10.14% showed bilateral volume differences that exceeded 10%. The study demonstrated the potential of the proposed method in high-throughput anatomical analysis and connectomics construction of the Drosophila brain. [ABSTRACT FROM AUTHOR]

The pollination process of kiwifruit flowers plays a crucial role in kiwifruit yield. Achieving accurate and rapid identification of the four stages of kiwifruit flowers is essential for enhancing pollination efficiency. In this study, to improve the efficiency of kiwifruit pollination, we propose a novel full-stage kiwifruit flower pollination detection algorithm named KIWI-YOLO, based on the fusion of frequency-domain features. Our algorithm leverages frequency-domain and spatial-domain information to improve recognition of contour-detailed features and integrates decision-making with contextual information. Additionally, we incorporate the Bi-Level Routing Attention (BRA) mechanism with C3 to enhance the algorithm's focus on critical areas, resulting in accurate, lightweight, and fast detection. The algorithm achieves a mAP0:5 of 91.6% with only 1.8M parameters, the AP of the Female class and the Male class reaches 95% and 93.5%, which is an improvement of 3.8%, 1.2%, and 6.2% compared with the original algorithm. Furthermore, the Recall and F1-score of the algorithm are enhanced by 5.5% and 3.1%, respectively. Moreover, our model demonstrates significant advantages in detection speed, taking only 0.016s to process an image. The experimental results show that the algorithmic model proposed in this study can better assist the pollination of kiwifruit in the process of precision agriculture production and help the development of the kiwifruit industry. [ABSTRACT FROM AUTHOR]

Due to the constrained performance of embedded chips in devices such as drones, real-time processing of simultaneous vehicle detection and road segmentation networks becomes challenging, leading to a lack of associative feature learning. To tackle these issues, we introduce a novel multi-task model for vehicle detection and road segmentation in unmanned aerial vehicle(UAV) Aerial Imagery. Our approach introduces a lightweight Ghost-Dilated convolution, combining the large receptive field of dilated convolution with the efficiency of Ghost convolution, resulting in fewer parameters and reduced computational load. Building upon this, we propose the Ghost-Atrous Spatial Pyramid Pooling (G-ASPP) module, a multi-scale feature extraction module that enhances the model's multi-scale characteristics while minimizing the increase in network parameters and computational requirements associated with Atrous Spatial Pyramid Pooling(ASPP) modules. The constructed multi-task UAV aerial vehicle detection and road segmentation network incorporates a carefully designed backbone, neck, detection head, and segmentation head. By refining existing lightweight backbone networks, our model achieves superior real-time performance and accuracy, demonstrating enhanced detection and segmentation accuracy with lower parameters and computational overhead. Experimental validation on a self-constructed multi-task dataset highlights the proposed model's improved segmentation and detection performance, particularly for small targets and narrow roads, confirming its effectiveness. This research contributes valuable insights to the study of multi-task networks in the realm of UAV vision. [ABSTRACT FROM AUTHOR]

(1) Background: Small objects in Unmanned Aerial Vehicle (UAV) images are often scattered throughout various regions of the image, such as the corners, and may be blocked by larger objects, as well as susceptible to image noise. Moreover, due to their small size, these objects occupy a limited area in the image, resulting in a scarcity of effective features for detection. (2) Methods: To address the detection of small objects in UAV imagery, we introduce a novel algorithm called High-Resolution Feature Pyramid Network Mamba-Based YOLO (HRMamba-YOLO). This algorithm leverages the strengths of a High-Resolution Network (HRNet), EfficientVMamba, and YOLOv8, integrating a Double Spatial Pyramid Pooling (Double SPP) module, an Efficient Mamba Module (EMM), and a Fusion Mamba Module (FMM) to enhance feature extraction and capture contextual information. Additionally, a new Multi-Scale Feature Fusion Network, High-Resolution Feature Pyramid Network (HRFPN), and FMM improved feature interactions and enhanced the performance of small object detection. (3) Results: For the VisDroneDET dataset, the proposed algorithm achieved a 4.4% higher Mean Average Precision (mAP) compared to YOLOv8-m. The experimental results showed that HRMamba achieved a mAP of 37.1%, surpassing YOLOv8-m by 3.8% (Dota1.5 dataset). For the UCAS_AOD dataset and the DIOR dataset, our model had a mAP 1.5% and 0.3% higher than the YOLOv8-m model, respectively. To be fair, all the models were trained without a pre-trained model. (4) Conclusions: This study not only highlights the exceptional performance and efficiency of HRMamba-YOLO in small object detection tasks but also provides innovative solutions and valuable insights for future research. [ABSTRACT FROM AUTHOR]

The underwater light scattering, absorption, and camera or target moving often bring issues such as blurring, distortion, and color deviation in underwater imaging, which poses significant challenges to underwater target detection. Numerous detectors have been proposed to address these challenges, such as YOLO series models, RCNN-based variants, and Transformer-based variants. However, the previous detectors often have poor detection results when encountering small targets and target occlusion problems. To tackle these issues, We propose a feature fusion and global semantic decoupling head-based YOLO detection method. Specifically, we propose an efficient feature fusion module to solve the problem of small target feature information being lost and difficult to detect accurately. We also use self-supervision to recalibrate the feature information between each level, which achieves full integration of semantic information between different levels. We design a decoupling head that focuses on global context information, which can better filter out complex background information, thereby achieving effective detection of targets under occluded backgrounds. Finally, we replace simple upsampling with a content-aware reassembly module in the YOLO backbone, alleviating the problem of imprecise localization and identification of small targets caused by feature loss to some extent. The experimental results indicate that the proposed method achieves superior performance compared to other state-of-the-art single-stage and two-stage detection networks. Specifically, on the UTDAC2020 dataset, the proposed method attains mAP50-95 and mAP50 scores of 54.4% and 87.7%, respectively. [ABSTRACT FROM AUTHOR]

The rapid evolution of deep learning techniques, applied in the smart cities context, has revolutionized computer vision applications, with particular significance in the field of object detection. In this paper, we explore the application of deep learning for real-time car detection in visual data, such as images or video streams. The proposed deep learning model is trained on large-scale datasets containing diverse images of cars, encompassing various lighting conditions, weather patterns, and traffic scenarios. Moreover, once the presence of car(s) is detected, we consider finding the distance between the detected car(s) and the camera: in this way, it is possible to understand the distance to the vehicle to take a certain countermeasure (for example braking or slowing down). The experimental analysis, performed on a dataset composed of 25532 different images confirms the effectiveness of the proposed method for real-world car detection. Moreover, examples of the distance computed between the camera and the detected cars are shown to provide samples of the adoption of the proposed method in the real-world environment. [ABSTRACT FROM AUTHOR]

The accurate location of vocal cords is crucial in medical imaging applications for diagnostic and therapeutic purposes. In this study, we evaluated the performance of state-of-the-art object detection models, including the YOLO, EfficientDet D1, Faster R-CNN and Single Shot Detector architectures, for the task of vocal cord detection. Using a data set of annotated medical images, we conducted comprehensive experiments to compare the models in terms of mean average precision (mAP) and inference time. Our results indicate that YOLOv8m achieved the highest performance, with a mAP@[0.5:0.95] of 0.778 and an inference time of 4.7 milliseconds. EfficientDet D1, Faster R-CNNs, and Single Shot Detectors trailed behind with slightly lower mAP scores. These findings highlight the efficacy of YOLOv8m for vocal cord localization on the utilized dataset. In addition, our study provides valuable information for the selection of suitable object detection models in medical imaging applications, facilitating better diagnosis and treatment planning in laryngology and otolaryngology. [ABSTRACT FROM AUTHOR]

This study presents a comparative analysis of segmentation models based on the YOLO (You Only Look Once) architecture for the task of vocal cord detection in laryngoscopic images. The yolov5, yolov8, and yolov9 architectures were evaluated using images obtained from laryngoscopic videos recorded during standard examinations at ORL clinics. The primary objective was to assess the efficiency of different model sizes and architectures in accurately identifying the position of vocal cords within the images. Our findings reveal that all evaluated architectures demonstrate proficiency in vocal cord detection, with comparable results across the models. However, there is a discernible difference in mean Average Precision (mAP) metrics (at IoU thresholds ranging from 0.5 to 0.95). Notably, yolov8 exhibits the highest mAP scores, followed by yolov5 and yolov9, indicating superior performance in identifying vocal cord regions. This comparative analysis provides valuable insights into the effectiveness of YOLO-based segmentation models for vocal cord detection, highlighting the importance of model size and architecture selection in medical image analysis applications. [ABSTRACT FROM AUTHOR]

AI analytics enables autonomous cars to detect and recognize objects, such as other vehicles, pedestrians, traffic signs, and obsta- cles, in real-time. Deep learning models, notably the You Only Look Once (YOLO) model, have demonstrated accuracy and speed in obstacle avoidance. However, current datasets are limited, lacking diversity and labeling, hindering their ability to represent real- world scenarios accurately. Besides, previous studies have focused extensively on specific object classes, such as pedestrians and vehicles, often neglecting other objects like bikes and road signs. To address this, we introduce a novel dataset tailored for AV envi- ronments, encompassing various road object types under different conditions. Our innovative methodology relies on self-supervised learning using the late YOLO version to improve model robustness with limited labeled data and AI-driven adaptive model opti- mization based on real-time feedback. We evaluate three YOLO architectures—YOLOv5, YOLOv7, and YOLOv8—customized for AV object detection. Our assessment covers everyday AV objects such as cars, pedestrians, bicycles, and road signs, empha- sizing early detection. We employ the VSim-AV simulator dataset to ensure robust evaluation, augmented with preprocessing techniques to optimize data quality and model generalization. The study reveals that YOLOv5 and YOLOv8 outperform YOLOv7 regarding precision and recall across various object classes, with YOLOv5 leading at 1.3 ms/image and YOLOv8 at 3.3 ms/image. The mean average precision was 0.94 for YOLOv5, 0.441 for YOLOv7, and 0.927 for YOLOv8, highlighting the limitations in current literature and challenges in YOLO model performance. [ABSTRACT FROM AUTHOR]

The detection of road signs is pivotal for effective traffic management and the operation of autonomous vehicles. In this context, deep learning, especially with respect to object detection models, can make a substantial contribution by facilitating the direct detection of road signs from images captured by moving vehicles. In this paper, we propose a method aimed at detecting road signs by exploiting the You Only Look Once model. We experiment with a dataset composed by 10000 different road sign images by obtaining interesting performances, thus showing that the proposed model can be considered reliable for real-time road sign detection. [ABSTRACT FROM AUTHOR]

Deep Learning has the ability to train on datasets created from videos. This facility makes deep learning algorithms suitable for detecting distinct objects in large sets of frames, particularly for delineating anomalies like precancerous lesions during surveillance colonoscopies of the large bowel. However, capturing subtle, diffuse characteristics in these videos' frames can be a challenge. This paper presents a deep learning system that uses colour channel separation and recombination of images to improve the performance of an object detection model, to tackle this challenge. Using a dataset from surveillance colonoscopy videos to find precancerous and cancerous lesions in IBD patients, individual colour channels of RGB images in the dataset are separated and recombined to form different datasets which are later used to train YOLOv8x models. The object detection model that is trained and tested with only the blue channel component of the images in the dataset performs better and gives more accurate predictions than the object detection model that is trained and tested with the datasets containing the green and/or the red channel images, as well as the dataset containing the original RGB images. [ABSTRACT FROM AUTHOR]

The state of object detection has quickly progressed due to the rapid development of hardware and deep learning models for the said task. The continuous development of the latter allowed its application in real-time. That is, from two-stage to one-stage detectors, which greatly reduced the processing time; and from anchor-based to anchor-free detectors, which allowed a more efficient and flexible model implementation. The purpose of this study is to explore the various YOLO versions available in Ultralytics to detect large office objects for obstacle avoidance. More specifically, this study observed how the versions, 3u, 5mu, 8 and 6m, from which the first three are anchor-free, perform over the dataset. Based from the results, it was found that there were no significant differences between the detectors, except for the anchor-based v6m, which has the worst scores. The four models, v8n, v8s, v8m, and v5mu, were, however, chosen as the candidate models due to their processing speed, ranging from 66 to 75 FPS on the NVIDIA GeForce RTX 3060 GPU. The said models were also subjected to an optimization process in terms of the confidence and IoU threshold values to find the pair that could provide the best metric scores. Results have provided candidate models for the study's detection task for obstacle avoidance in real-time. [ABSTRACT FROM AUTHOR]

Effective safety management is vital for ensuring construction safety. Traditional safety inspections in construction heavily rely on manual labor, which is both time-consuming and labor-intensive. Extensive research has been conducted integrating computer-vision technologies to facilitate intelligent surveillance and improve safety measures. However, existing research predominantly focuses on singular tasks, while construction environments necessitate comprehensive analysis. This study introduces a multi-task computer vision technology approach for the enhanced monitoring of construction safety. The process begins with the collection and processing of multi-source video surveillance data. Subsequently, YOLOv8, a deep learning-based computer vision model, is adapted to meet specific task requirements by modifying the head component of the framework. This adaptation enables efficient detection and segmentation of construction elements, as well as the estimation of person and machine poses. Moreover, a tracking algorithm integrates these capabilities to continuously monitor detected elements, thereby facilitating the proactive identification of unsafe practices on construction sites. This paper also presents a novel Integrated Excavator Pose (IEP) dataset designed to address the common challenges associated with different single datasets, thereby ensuring accurate detection and robust application in practical scenarios. [ABSTRACT FROM AUTHOR]

Surface defect inspection methods have proven effective in addressing casting quality control tasks. However, traditional inspection methods often struggle to achieve high-precision detection of surface defects in castings with similar characteristics and minor scales. The study introduces DES-YOLO, a novel real-time method for detecting castings' surface defects. In the DES-YOLO model, we incorporate the DSC-Darknet backbone network and global attention mechanism (GAM) module to enhance the identification of defect target features. These additions are essential for overcoming the challenge posed by the high similarity among defect characteristics, such as shrinkage holes and slag holes, which can result in decreased detection accuracy. An enhanced pyramid pooling module is also introduced to improve feature representation for small defective parts through multi-layer pooling. We integrate Slim-Neck and SIoU bounding box regression loss functions for real-time detection in actual production scenarios. These functions reduce memory overhead and enable real-time detection of surface defects in castings. Experimental findings demonstrate that the DES-YOLO model achieves a mean average precision (mAP) of 92.6% on the CSD-DET dataset and a single-image inference speed of 3.9 milliseconds. The proposed method proves capable of swiftly and accurately accomplishing real-time detection of surface defects in castings. [ABSTRACT FROM AUTHOR]

Synthetic aperture radar (SAR) is a technique widely used in the field of ship detection. However, due to the high ship density, fore-ground-background imbalance, and varying target sizes, achieving lightweight and high-precision multiscale ship object detection remains a significant challenge. In response to these challenges, this research presents YOLO-MSD, a multiscale SAR ship detection method. Firstly, we propose a Deep Poly Kernel Backbone Network (DPK-Net) that utilizes the Optimized Convolution (OC) Module to reduce data redundancy and the Poly Kernel (PK) Module to improve the feature extraction capability and scale adaptability. Secondly, we design a BiLevel Spatial Attention Module (BSAM), which consists of the BiLevel Routing Attention (BRA) and the Spatial Attention Module. The BRA is first utilized to capture global information. Then, the Spatial Attention Module is used to improve the network's ability to localize the target and capture high-quality detailed information. Finally, we adopt a Powerful-IoU (P-IoU) loss function, which can adjust to the ship size adaptively, effectively guiding the anchor box to achieve faster and more accurate detection. Using HRSID and SSDD as experimental datasets, mAP of 90.2% and 98.8% are achieved, respectively, outperforming the baseline by 5.9% and 6.2% with a model size of 12.3 M. Furthermore, the network exhibits excellent performance across various ship scales. [ABSTRACT FROM AUTHOR]

This study addresses marine pollution caused by debris entering the ocean through rivers. A physical and bubble barrier system has been developed to collect debris, but an effective identification and classification system for incoming vessels is needed. This study evaluates the effectiveness of deep learning models in identifying and classifying vessels in real time. The YOLO (You Only Look Once) v5 and v8 models are evaluated for vessel detection and classification. A dataset of 624 images representing 13 different types of vessels was created to train the models. The YOLOv8, featuring a new backbone network, outperformed the YOLOv5 model, achieving a high mean average precision (mAP@50) of 98.9% and an F1 score of 91.6%. However, YOLOv8's GPU consumption increased by 116% compared to YOLOv5. The advantage of the proposed method is evident in the precision–confidence curve (PCC), where the accuracy peaks at 1.00 and 0.937 confidence, and in the achieved frames per second (fps) value of 84.7. These findings have significant implications for the development and deployment of real-time marine pollution control technologies. This study demonstrates that YOLOv8, with its advanced backbone network, significantly improves vessel detection and classification performance over YOLOv5, albeit with higher GPU consumption. The high accuracy and efficiency of YOLOv8 make it a promising candidate for integration into marine pollution control systems, enabling real-time identification and monitoring of vessels. This advancement is crucial for enhancing the effectiveness of debris collection systems and mitigating marine pollution, highlighting the potential for deep learning models to contribute to environmental preservation efforts. [ABSTRACT FROM AUTHOR]

Lunar exploration is pivotal in establishing a human presence on the Moon, and lunar crater detection plays a major role in this pursuit. The study is divided into two key phases: the creation of a specialized annotated dataset sourced from the Optical High-Resolution Camera on the Chandrayaan-2 satellite, and the evaluation of model performance using this dataset. Employing models such as FasterRCNN, YoloV5, and YoloV1, the investigation reveals the YoloV5 model's superiority, achieving a precision of 92% and a recall of 83% for lunar crater detection. This finding constitutes a significant contribution to lunar exploration research. [ABSTRACT FROM AUTHOR]

This study develops forest road recognition technology using deep learning-based image processing to support the advancement of autonomous driving technology for forestry machinery. Images were collected while driving a tracked forwarder along approximately 1.2 km of forest roads. A total of 633 images were acquired, with 533 used for the training and validation sets, and the remaining 100 for the test set. The YOLOv8 segmentation technique was employed as the deep learning model, leveraging transfer learning to reduce training time and improve model performance. The evaluation demonstrates strong model performance with a precision of 0.966, a recall of 0.917, an F1 score of 0.941, and a mean average precision (mAP) of 0.963. Additionally, an image-based algorithm is developed to extract the center from the forest road areas detected by YOLOv8 segmentation. This algorithm detects the coordinates of the road edges through RGB filtering, grayscale conversion, binarization, and histogram analysis, subsequently calculating the center of the road from these coordinates. This study demonstrates the feasibility of autonomous forestry machines and emphasizes the critical need to develop forest road recognition technology that functions in diverse environments. The results can serve as important foundational data for the future development of image processing-based autonomous forestry machines. [ABSTRACT FROM AUTHOR]

Object detection is a computer vision task that identifies and locates one or more effective targets from image or video data. The accuracy of object detection heavily depends on the size and the diversity of the utilized dataset. However, preparing and labeling an adequate dataset to guarantee a high level of reliability can be time-consuming and labor-intensive, because the process of building data requires manually setting up the environment and capturing the dataset while keeping its variety in scenarios. There have been several efforts on object detection that take a long time to prepare the input data for training the models. To deal with this problem, synthetic data have emerged as a potential for the replacement of real-world data in data preparation for model training. In this paper, we provide a technique that can generate an enormous synthetic dataset with little human labor. Concretely, we have simulated the environment by applying the pyBullet library and capturing various types of input images. In order to examine its performance on the training model, we integrated a YOLOv5 object detection model to investigate the dataset. The output of the conducted model was deployed in a simulation robot system to examine its potential. YOLOv5 can reach a high accuracy of object detection at 93.1% mAP when solely training on our generated data. Our research provides a novelistic method to facilitate the understanding of the data generation process in preparing datasets for deep learning models. [ABSTRACT FROM AUTHOR]

Nowadays, children face different changes and challenges from an early age, which can have long-lasting impacts on them. Many children struggle to express or explain their feelings and thoughts properly. Due to that fact, psychological and mental health specialists found a way to detect mental issues by observing and analyzing different signs in children’s drawings. Yet, this process remains complex and time-consuming. This study proposes a solution by employing artificial intelligence to analyze children’s drawings and provide diagnosis rates with high accuracy. While prior research has focused on detecting psychological and mental issues through questionnaires, only one study has explored analyzing emotions in children's drawings by detecting positive and negative feelings. A notable gap is the limited diagnosis of specific mental issues, along with the promising accuracy of the detection results. In this study, different versions of YOLO were trained on a dataset of 500 drawings, split into 80% for training, 10% for validation, and 10% for testing. Each drawing was annotated with one or more emotional labels: happy, sad, anxiety, anger, and aggression. YOLOv8-cls, YOLOv9, and ResNet50 were used for object detection and classification, achieving accuracies of 94%, 95.1%, and 70.3%, respectively. YOLOv9 and ResNet50 results were obtained at high epoch numbers with large model sizes of 5.26 MB and 94.3 MB. YOLOv8-cls achieved the most satisfying result, reaching a high accuracy of 94% after 10 epochs with a compact model size of 2.83 MB, effectively meeting the study's goals. [ABSTRACT FROM AUTHOR]

A road intersection is an area where more than two roads in different directions connect. It is a point of transition where the driver navigates and makes the decision, making it an area with a high risk for traffic accidents. Road intersection detection is identifying and analyzing road intersections in real time using various technologies and algorithms. It is an essential part of intelligent transportation systems and autonomous driving. Road intersection detection helps the driver to identify the road intersection early to make good driving decisions and avoid accidents. Despite its high importance, only a few research is found regarding this topic. Existing research mainly focuses on detecting and classifying traffic signs, vehicles, and pedestrians. In this research, we propose an algorithm to detect road intersections using an image from the front-facing camera installed on the car as an input. We use traffic sign detection to detect seven types of traffic signs having a high probability of intersection nearby and combine it with our novel road intersection detection algorithm to detect the location of the road intersection. Our road inter-section detection algorithm leverages the relationship between the area of the traffic signs and the location of the intersection. Our proposed method gives promising results from the experiments and can detect road intersections from further distances. Our method is also able to perform detection in real time. [ABSTRACT FROM AUTHOR]

Roads are one of the most critical infrastructures, which should be maintained at a high quality of service. For this purpose, road pavement should be assessed cost-effectively. In the past, image processing methods were used to analyze pavement conditions. In recent years, machine learning methods have been employed, while now deep learning methods are applied. Deep learning has outperformed other methods regarding the accuracy and speed of pavement distress evaluation. In this research, a deep learning algorithm called YOLOv5 is deployed to detect pavement block cracking and estimate its severity using images taken from the right of way via a road surface profiler. Two models are successfully trained and tested, one to detect block cracking and the other to predict its severity with a sufficient level of accuracy of 84.5% and 76.6%, respectively. It is concluded that the model not only can detect block cracking but also predict its severity. [ABSTRACT FROM AUTHOR]

With the development of computer vision and remote sensor devices, object detection in aerial images has drawn considerable attention because of its ability to provide a wide field of view and a large amount of information. Despite this, object detection in aerial images is a challenging task owing to densely packed objects, oriented diversity, and complex background. In this study, we optimized three aspects of the YOLOv5 algorithm to detect arbitrary oriented objects in remote sensing images, including head structure, features from the backbone, and angle prediction. To improve the head structure, we decoupled it into four submodules, which are used for object localization, foreground, category, and oriented angle classification. To increase the accuracy of the features from the backbone, we designed a block dimensional attention module, which is developed by splitting the image into smaller patches based on a dimensional attention module. Compared with the original YOLOv5 algorithm, our approach has a better performance for oriented object detection-the mAP on DOTA-v1.5 is increased by 1.25%. It was tested to be effective on DOTA-vl.0, HRSC2016, and DIOR-R datasets as well. [ABSTRACT FROM AUTHOR]

In this study, we aim to validate the use of Artificial Intelligence of Things (AIoT) for deep learning-based detection of piglet crushing incidents by sows in pig farms, a leading cause of piglet mortality. To achieve this, we developed a deep neural network based on the You Only Look Once (YOLO) model, which was trained to detect piglet crushing events in real-time. We then optimized the weights of the YOLO model using TensorFlow Lite, TensorRT, and edge TPU to ensure efficient operation on AIoT devices with limited computational resources. Finally, we compared the performance of the deep learning-based piglet crushing detection algorithm on various single-board computers (SBCs), including the Jetson Nano, Raspberry Pi, and Coral Dev Board, to validate its suitability for real-world applications. [ABSTRACT FROM AUTHOR]

As urbanization continues to reshape transportation, the safety of cyclists in complex traffic environments has become a pressing concern. In response to this challenge, our research introduces a CycleInSight framework, which harnesses advanced deep learning and computer vision techniques to enable precise and efficient cyclist detection in diverse urban settings. Utilizing you only look once version 8 (YOLOv8) object detection algorithm, the proposed model aims to detect and localize vulnerable cyclists near vehicles equipped with onboard cameras. Our research presents comprehensive experimental results demonstrating its effectiveness in identifying vulnerable cyclists amidst dynamic and challenging traffic conditions. With an impressive average precision of 90.91%, our approach outperforms existing models while maintaining efficient inference speeds. By effectively identifying and tracking cyclists, this framework holds significant potential to enhance urban traffic safety, inform data-driven infrastructure planning, and support the development of advanced driver assistance systems and autonomous vehicles. [ABSTRACT FROM AUTHOR]

Detecting the objects and tracking them is a wonder with the help of the recent technologies. In today's life, video surveillance is commonly present at maximum places. As the population increases, the crime rates also keep on increasing at different cases. The crime identification, dealing with the crime scenes and tracking down the criminal is a major task which involves numerous man powers. If there is a methodology for detecting and preventing the crime, it would be very much helpful to the public and also to the authorities. So the main objective of the proposed system is to automate the crime identification and crime tracking process by using the video surveillance with deep learning algorithms. Also, to alert the public before a crime takes place in public areas. Because of the timely alert, further crime activities may be avoided or the loss incurred may be reduced. The system also helps in identifying the criminals and tracking the information about the crime scenes, thus providing more useful information to the authorities in solving the cases. According to recent research works, the YOLO algorithm achieves higher accuracy with multi-object detection. In this paper, a framework with improvised YOLO algorithm is illustrated. The algorithm is fine-tuned with different hyperparameters for achieving the AUC with 0.91 for detecting vandalism behavior and 0.8299 over all the 14 classes of crime activities. The result of the proposed system is compared with the existing systems with parameters like training loss, testing loss, precision and F1 score. [ABSTRACT FROM AUTHOR]

Icing detection promptly on all-dielectric self-supporting (ADSS) optical fiber cable on overhead transmission lines is very difficult, so it presents an icing detection method based on the improved YOLOv8 network in this paper. Initially, the improved YOLOv8 network is used to detect and extract the regions of ADSS optical fiber cable icing, reducing interference from complex background. The introduction of the GE attention module in the YOLOv8 network enhances its adaptability to various environments, and improves the recognition of icing states in different scenarios. The BiFPN feature fusion module is incorporated to replace the concatenate structure in the original YOLOv8 network, in order to improve detection accuracy. Subsequently, based on RCF edge detection algorithm, icing optical fiber cable edge feature is obtained in the RGB images. Finally, the icing thickness is computed based on pre-icing image. The experimental results showed that compared to the original YOLOv8 network, this improved network improves the accuracy of icing detection by 5.9%, the recall rate by 5%, and the average accuracy by 2.2%. The algorithm achieved the detection speed of 147 FPS, and an error rate of 2.5% in the icing thickness detection, which provided a valuable reference for ADSS optical fiber cable monitoring and ensuring the normal operation of the power signal network, as well as for early warning of icing on ADSS optical fiber cables. [ABSTRACT FROM AUTHOR]

Object detection in high resolution enables the identification and localization of objects for monitoring critical areas with precision. Although there have been improvements in object detection at high resolution, the variety of object scales, as well as the diversity of backgrounds and textures in high-resolution images, make it challenging for detectors to generalize successfully. This study introduces a new method for object detection in high-resolution images. The pre-processing stage of the method includes ISA and SAM to slice the input image and segment the objects in bounding boxes, respectively. In order to improve the resolution in the slices, the first layer of YOLO is designed as SRGAN. Thus, before applying YOLO detection, the resolution of the sliced images is increased to improve features. The proposed system is evaluated on xView and VisDrone datasets for object detection algorithms in satellite and aerial imagery contexts. The success of the algorithm is presented in four different YOLO architectures integrated with SRGAN. According to comparative evaluations, the proposed system with Yolov5 and Yolov8 produces the best results on xView and VisDrone datasets, respectively. Based on the comparisons with the literature, our proposed system produces better results. [ABSTRACT FROM AUTHOR]

Aiming at the problem of low accuracy of multi-scale seafloor target detection in side-scan sonar images with high noise and complex background texture, a model for multi-scale target detection using the BES-YOLO network is proposed. First, an efficient multi-scale attention (EMA) mechanism is used in the backbone of the YOLOv8 network, and a bi-directional feature pyramid network (Bifpn) is introduced to merge the information of different scales, finally, a Shape_IoU loss function is introduced to continuously optimize the model and improve its accuracy. Before training, the dataset is preprocessed using 2D discrete wavelet decomposition and reconstruction to enhance the robustness of the network. The experimental results show that 92.4% of the mean average accuracy at IoU of 0.5 (mAP@0.5) and 67.7% of the mean average accuracy at IoU of 0.5 to 0.95 (mAP@0.5:0.95) are achieved using the BES-YOLO network, which is an increase of 5.3% and 4.4% compared to the YOLOv8n model. The research results can effectively improve the detection accuracy and efficiency of multi-scale targets in side-scan sonar images, which can be applied to AUVs and other underwater platforms to implement intelligent detection of undersea targets. [ABSTRACT FROM AUTHOR]

Computer vision models such as You Only Look Once (YOLO) excel at tracking general objects, yet they often struggle to accurately track multiple animals. In this study, we aimed to verify the improvement of animal tracking performance using YOLO by fine-tuning. We obtained videos of multiple domestic cats by ourselves and annotated the videos, and performed fine-tuning with this video dataset. Results of this study show that even training with less than one hundred images improves accuracy and enables tracking of individual animals. This result suggests that our method is promising for the realization of animal social behavior analysis. [ABSTRACT FROM AUTHOR]

Potatoes have been a versatile food and cash crop to ensure food security and grain planting structure in China. The total planting area has declined in the planting industries at present. Some challenges also remain in the mechanization of the potato planting industry. The current process of seed potato cutting can rely heavily on manual and mechanical blind cutting, leading to labor-intensive and inefficient tasks. Moreover, there is a high rate of blind cuts and significant loss of seed potatoes. Therefore, it is highly urgent to accurately and rapidly identify the potato eyes before cutting. In this study, an improved model was proposed to detect the potato bud eyes using YOLOv5. Dutch 15 potato variety was taken as the experimental material. The high-quality samples of seed potatoes were carefully chosen to be free of diseases, dry rot, disease spots, and worm eyes. The dataset was used for training, verification, and testing. 1 400 pictures of seed potatoes were obtained with a ratio of 8:1:1. Data expansion techniques (such as mirroring, rotating, cropping, and adjusting brightness) were applied to enhance the dataset, thus resulting in a total of 5 600 images. Since the features of seed potato eyes were relatively simple, there was a decrease to even disappear after multiple convolutions. C3 Faster was integrated into the original framework. While the extraction of sprout eye features was enhanced to reduce the parameters. Additionally, the GD structure was incorporated from the Neck component of GOLD-YOLO to improve the detection accuracy of sprout eye. The bounding box loss function CIoU Loss was replaced with WIoU Loss to expedite the convergence of the network model for high detection accuracy. The hyperparameters of the original YOLO model were optimized for the COCO dataset. There were significant differences from the dataset of seed potato sprout eye in this experiment. Therefore, a genetic algorithm (GA) was employed to fine-tune the hyperparameters specifically for the detection of seed potato sprout eye at the end of the experiment. Furthermore, pruning and distillation techniques were used to reduce the running parameters and memory consumption, in order to make the model more suitable for the detection tasks of seed potato bud eye. The optimized model was presented with a size of 8.7 MB, which was only 61.3% of the original model. Params comprised approximately 57.1% of the original model. The final average accuracies of detection were 90.5% and 90.1%, respectively, in the test and validation sets of the self-made potato dataset. Compared with the lightweight networks YOLOv7-tiny, YOLOv8n, YOLOv5n, and YOLOv5s, the average accuracies of the improved model were 0.5, 1.3, 2.8, and 1.1 percentage points higher, respectively, in the test set of seed potato dataset. The average accuracy of the verification set was 2.9, 1.9, 3.2, and 1.6 percentage points higher. The detection speed on a local computer reached 27.5 frames per second, fully meeting the real-time requirements. Substantial benefits were offered to significantly enhance the efficiency and accuracy of seed potato bud eye detection in the potato cultivation industry. [ABSTRACT FROM AUTHOR]