Article
Green & Sustainable Science & Technology
Junlong Tang, Shenbo Liu, Dongxue Zhao, Lijun Tang, Wanghui Zou, Bin Zheng
Summary: In this paper, an improved detection algorithm of PCB surface defects based on YOLOv5, named PCB-YOLO, is proposed to address the problems of low network accuracy, slow speed, and a large number of model parameters in PCB defect detection. The algorithm obtains more suitable anchors for the dataset using the K-means++ algorithm and adds a small target detection layer to focus on more small target information. Swin transformer is embedded into the backbone network, and a united attention mechanism is constructed to improve the network's analysis ability. Model volume compression is achieved by introducing depth-wise separable convolution. The EIoU loss function is used to optimize the regression process and enhance the localization ability of small targets. Experimental results show that PCB-YOLO achieves a satisfactory balance between performance and consumption, with 95.97% mAP at 92.5 FPS, making it more accurate and faster than many other algorithms for real-time and high-precision detection of product surface defects.
Article
Engineering, Electrical & Electronic
Huibin Shi, Zexi Hua, Jianyi Chen, Yongchuan Tang, Rujiang He
Summary: This paper proposes a detection and identification method based on the improved YOLO-v3 for substation digital instruments, aiming to monitor substations intelligently. By augmenting the image dataset and using the PANet structure, the proposed method achieves accurate detection and identification of substation instruments with real-time performance, meeting the actual needs of substation data acquisition and engineering.
SIGNAL IMAGE AND VIDEO PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Qirong Zhang, Zhong Han, Yu Zhang
Summary: This article proposes an improved YOLO V3 network model to address the issue of insufficient small target detection ability in existing network models. The algorithm model improvement effectively enhances the detection ability of small target vehicles in aerial photography. The optimization of anchor boxes, adjustment of the network residual module, and introduction of rectangular prediction frames with orientation angles have improved the algorithm's small target detection effect and vehicle positioning efficiency.
PEERJ COMPUTER SCIENCE
(2023)
Article
Engineering, Marine
Yue Li, Xueting Zhang, Zhangyi Shen
Summary: Due to the strain on land resources, marine energy development is expanding, and as a result, submarine cable inspections are required. This research proposes an improved YOLO-SC detection method and demonstrates its effectiveness in submarine cable detection through experiments.
JOURNAL OF MARINE SCIENCE AND ENGINEERING
(2022)
Article
Computer Science, Artificial Intelligence
Ke Zhao, Yaonan Wang, Yi Zuo, Chujin Zhang
Summary: This paper proposes an improved method for detecting positioning bolts (PB) by enhancing the dataset, designing an improved anchor box mechanism, and enhancing the feature extraction network. The method improves the detection accuracy and speed of PB data in palletizing robots.
JOURNAL OF INTELLIGENT & ROBOTIC SYSTEMS
(2022)
Article
Computer Science, Information Systems
Kshitij Dhawan, R. Srinivasa Perumal, R. K. Nadesh
Summary: The ability of Advanced Driving Assistance Systems (ADAS) to identify and understand all objects around the vehicle is critical. Today's vehicles are equipped with advanced driving assistance systems that make driving safer and more comfortable. A camera helps the system recognize and detect traffic signs, and algorithms are used to classify them with high accuracy.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Cheng Xianbao, Qiu Guihua, Jiang Yu, Zhu Zhaomin
Summary: This paper proposes an improved algorithm based on Yolo V3 to enhance small target detection accuracy through three optimizations: improving the feature map acquisition network, adding a size recognition module, and enhancing feature channel. The new algorithm improves the detection accuracy, recall rate, and average accuracy of small objects compared to Yolo V3.
PATTERN ANALYSIS AND APPLICATIONS
(2021)
Article
Plant Sciences
Jun Liu, Xuewei Wang, Wenqing Miao, Guoxu Liu
Summary: This research proposes a tomato pest identification algorithm based on an improved YOLOv4 fusing triplet attention mechanism, addressing the issue of imbalances in sample numbers and demonstrating high recognition accuracy in experiments. The algorithm's performance on practical images also supports its feasibility for tomato pest detection.
FRONTIERS IN PLANT SCIENCE
(2022)
Article
Mathematical & Computational Biology
Zeyong Huang, Yuhong Li, Tingting Zhao, Peng Ying, Ying Fan, Jun Li
Summary: This paper proposes a liquid level detection model based on deep learning to improve the accuracy of intravenous infusion liquid level detection, reduce patient pain, and demonstrates high precision and real-time performance in experiments.
MATHEMATICAL BIOSCIENCES AND ENGINEERING
(2021)
Article
Computer Science, Information Systems
Zhonglai Yang
Summary: This paper improves the YOLO v3 algorithm to achieve road traffic sign recognition, and achieves better accuracy and speed by enhancing the data and optimizing the network structure.
MOBILE INFORMATION SYSTEMS
(2022)
Article
Materials Science, Multidisciplinary
Xiaoning Cui, Qicai Wang, Jinpeng Dai, Rongling Zhang, Sheng Li
Summary: The study focused on intelligent identification of erosion damage to concrete using deep learning, establishing a dataset and proposing an improved YOLO-v3 algorithm model. The improved YOLO-v3 algorithm shows more accurate recognition of concrete erosion damage compared to other mainstream target detection algorithms, with an accuracy, precision, and MAP of 96.32%, 95.68%, and 75.68% respectively, validating the applicability of deep learning in the research of concrete erosion damage.
Article
Computer Science, Artificial Intelligence
Hongguang Pan, Yuhong Shi, Xinyu Lei, Zheng Wang, Fangfang Xin
Summary: This paper proposes an improved tiny-YOLO-v3-based fast identification model for efficient gangue sorting.
JOURNAL OF REAL-TIME IMAGE PROCESSING
(2022)
Article
Environmental Sciences
Jinrui Li, Libin Chen, Jian Shen, Xiongwu Xiao, Xiaosong Liu, Xin Sun, Xiao Wang, Deren Li
Summary: Fast and high-accuracy detection of underwater targets based on side scan sonar images has great potential for various applications. An improved neural network with spatial pyramid pooling and online dataset preprocessing based on YOLO V3 is proposed to address the limitations of low-resolution images and improve detection performance. Experimental results show that the proposed method achieves higher detection accuracy compared to other algorithms.
Article
Computer Science, Artificial Intelligence
Tu Renwei, Zhu Zhongjie, Bai Yongqiang, Gao Ming, Ge Zhifeng
Summary: This paper proposes a deep learning detection model based on YOLO v3, which improves detection accuracy, flops, and speed by simplifying the neural network structure and optimizing the dataset during training.
INTERNATIONAL ARAB JOURNAL OF INFORMATION TECHNOLOGY
(2021)
Article
Agronomy
Lifa Fang, Yanqiang Wu, Yuhua Li, Hongen Guo, Hua Zhang, Xiaoyu Wang, Rui Xi, Jialin Hou
Summary: This study utilizes object detection techniques in deep learning to propose a ginger recognition network based on YOLOv4-LITE, which successfully solves the issue of consistent orientation in ginger planting, providing a technical guarantee for automated seeding.
