Article
Ecology
Hongtao Zheng, Junchen Duan, Yu Dong, Yan Liu
Summary: This paper proposes a real-time fire detection algorithm based on MobileNetV3-large and yolov4, which achieves high recognition accuracy and real-time performance. By replacing CSP Darknet53 in yolov4 with MobileNetV3-large, the algorithm reduces the computational effort of the network structure while extracting flame and smoke features. Furthermore, the algorithm is improved by adding PANet, SPP, Vision Transformer, and ECA Net components to enhance feature extraction capability and recognition performance.
Article
Plant Sciences
Lele Wang, Yingjie Zhao, Shengbo Liu, Yuanhong Li, Shengde Chen, Yubin Lan
Summary: This paper proposes a lightweight model based on the improved YOLOv4 to detect dense plums in orchards. By employing data augmentation and optimizing feature extraction, the improved model achieves better performance in terms of accuracy and speed. Experimental results show that the improved model outperforms other models in plum detection, demonstrating high accuracy and robustness.
FRONTIERS IN PLANT SCIENCE
(2022)
Article
Computer Science, Information Systems
Jiawei Zhao, Guangzhao Tian, Chang Qiu, Baoxing Gu, Kui Zheng, Qin Liu
Summary: This study proposes an improved YOLOv4 model for weed detection in potato fields. The model utilizes a lightweight network structure and introduces new convolution methods to improve detection speed and accuracy. Various image processing techniques and optimization algorithms are applied to enhance the robustness of the model. Experimental results show that the proposed method outperforms other commonly used models in terms of performance and effectiveness.
Article
Chemistry, Multidisciplinary
Xinting Liao, Shengping Lv, Denghui Li, Yong Luo, Zichun Zhu, Cheng Jiang
Summary: The improved YOLOv4 algorithm for PCB surface defect detection achieves higher accuracy and faster speed with lower memory consumption and fewer multiply-accumulate operations compared to the cutting-edge YOLOv4. The detector performed well in experiments using a customized dataset, outperforming other state-of-the-art detectors.
APPLIED SCIENCES-BASEL
(2021)
Article
Computer Science, Artificial Intelligence
Jun Tang, Zidong Wang, Hongyi Zhang, Han Li, Peishu Wu, Nianyin Zeng
Summary: This paper proposes a lightweight printed circuit board defects detection model (light-PDD) to overcome the deficiencies of redundant parameters and slow inference speed in existing methods. The light-PDD model follows the overall framework of YOLOv4 with enhancements, using a pruned MobileNetV3 structure for feature extraction. It also incorporates a dual-domain attention mechanism and diverse activation functions to effectively handle the detection of tiny-size PCB defects. The improved cross-stage partial structure is deployed for feature fusion to reduce model complexity. Experimental results demonstrate the superiority of light-PDD in terms of inference speed and detection accuracy.
EXPERT SYSTEMS WITH APPLICATIONS
(2024)
Article
Chemistry, Multidisciplinary
Jingxin Fang, Xuwei Li
Summary: This paper proposes a method for detecting irregular behaviors of substation personnel based on an improved YOLOv4 algorithm. The experimental results show that the method has high accuracy and fast detection speed, outperforming other object detection methods.
APPLIED SCIENCES-BASEL
(2022)
Article
Physics, Multidisciplinary
Tangbo Bai, Jialin Gao, Jianwei Yang, Dechen Yao
Summary: Traditional machine vision methods are inadequate for the detection of railway surface defects, hence this paper proposes a new method based on an improved YOLOv4, which achieves lightweight network and real-time detection, significantly improving detection accuracy.
Article
Computer Science, Information Systems
Chang Qiu, Guangzhao Tian, Jiawei Zhao, Qin Liu, Shangjie Xie, Kui Zheng
Summary: A grape maturity detection and visual pre-positioning algorithm based on improved YOLOv4 was proposed in this study to guide grape picking robots in recognizing and classifying grapes with different maturity levels in complex orchard environments. The algorithm achieved high accuracy and detection speed, with an overall average accuracy of 93.52% and an average detection time of 10.82 ms. The pre-positioning method based on binocular stereo vision obtained spatial position information of grape clusters with low error rates, making it reliable for precise grape picking in orchard environments.
Article
Forestry
Ao Li, Yaqin Zhao, Zhaoxiang Zheng
Summary: This paper proposes a model based on RBiFPN and Swin Transformer for detecting wildfire smoke. The model fuses and enhances smoke features, allowing for better differentiation between clouds and smoke, and achieves higher detection performance in most cases.
Article
Engineering, Multidisciplinary
D. L. Yuan, Y. Xu
Summary: By replacing the backbone network and utilizing deep separable convolution, this study successfully optimized the vehicle detection model, improving accuracy and reducing parameter count, laying the foundation for intelligent transportation systems.
ENGINEERING LETTERS
(2021)
Article
Engineering, Electrical & Electronic
Limei Song, Jiawei Kang, Qile Zhang, Shuopeng Wang
Summary: This paper proposes a lightweight detector called Light-YOLO-Welding based on an improved YOLOv4 for real-time recognition of weld feature locations in compounded noises. By modifying the backbone network and using bidirectional feature pyramid network, the method achieves significant improvement in detection speed and accuracy. Experimental results show that this method has higher mAP and accuracy, and is more reliable and efficient compared to other methods.
SIGNAL IMAGE AND VIDEO PROCESSING
(2023)
Article
Environmental Sciences
Fuzhen Zhu, Yuying Wang, Jingyi Cui, Guoxin Liu, Huiling Li
Summary: To address the issues of false detection, inadequate regression performance of anchor frames, and the inability to detect small targets in traditional multiscale target detection methods based on YOLOv4, we propose Enhanced YOLOv4. Our approach incorporates an improved BiFPN for multi-scale feature fusion, a channel attention mechanism (CAM) to highlight correlation between channels for small target detection, and a modified net training loss function with CDIoU for better anchor box regression and training speed. Experimental results on the DOTA dataset show that our method achieves a mAP of 90.88% and a frame rate of 58.76 FPS, with no significant impact on detection speed.
EGYPTIAN JOURNAL OF REMOTE SENSING AND SPACE SCIENCES
(2023)
Article
Plant Sciences
Zebing Zhang, Dapeng Jiang, Huiling Yu, Yizhuo Zhang
Summary: This study designed a machine vision detection model (EBE-YOLOV4) to achieve rapid and accurate recognition of pine cones in the forest. By optimizing the backbone and neck networks, the model achieved good measurement accuracy and detection speed.
FRONTIERS IN PLANT SCIENCE
(2022)
Article
Environmental Sciences
Yuchao Zhu, Jun Zhou, Yinhui Yang, Lijuan Liu, Fei Liu, Wenwen Kong
Summary: The study proposes an improved YOLOv4 model, which combines Mobilenetv3 network, CBAM module, and ASFF module, and optimizes the detection and counting of fruit tree canopies using the K-means algorithm, linear scaling, and cosine annealing learning strategy. The results show that the improved model can achieve fast and accurate recognition and counting of fruit tree canopies in orchard environments, with high detection accuracy and counting precision.
Article
Computer Science, Artificial Intelligence
Shengying Wang, Jing Zhao, Na Ta, Xiaoye Zhao, Mingxia Xiao, Haicheng Wei
Summary: This study proposes a deep learning fire recognition algorithm based on model compression and lightweight requirements, utilizing the MobileNetV3 model to simplify the conventional network structure and improve detection accuracy through knowledge distillation. Experimental results demonstrate significant advantages in reducing model parameters and inference time compared to existing algorithms.
JOURNAL OF REAL-TIME IMAGE PROCESSING
(2021)