Article
Multidisciplinary Sciences
Chao Liang, Zhengang Yan, Meng Ren, Jiangpeng Wu, Liping Tian, Xuan Guo, Jie Li
Summary: In this paper, a YOLOv5s-THSE model is proposed based on the YOLOv5s model, which extracts deeper target features by adding a multi-head attention mechanism to the backbone and neck of the network. The Cross Stage Partial, Squeeze-and-Exclusion module is added to suppress complex backgrounds, and a small object detection head is introduced with the CIoU loss function for improved detection accuracy. Experimental results show that the proposed model can effectively improve the detection performance of infrared tank targets under ground background compared to existing methods.
SCIENTIFIC REPORTS
(2023)
Article
Computer Science, Information Systems
Naeem Ayoub, Peter Schneider-Kamp
Summary: The use of deep learning-based autonomous drone vision systems shows promising results in detecting faults in power line components, providing an effective solution for real-time on-board power line inspection. Various single-board devices were utilized for experimental evaluation in running deep learning models.
Article
Computer Science, Artificial Intelligence
Zuyi Wang, Wei Zhao, Li Xu
Summary: The authors propose a lightweight sub-network called Quality Prediction Block (QPB) to enhance the effect of localization quality evaluation on detection confidence in one-stage object detectors. Extensive experiments on public benchmarks demonstrate the effectiveness of their method and its better performance in stronger one-stage detectors.
IET COMPUTER VISION
(2023)
Article
Engineering, Electrical & Electronic
Xiaoyu Chen, Hongliang Li, Qingbo Wu, King Ngi Ngan, Linfeng Xu
Summary: This paper introduces PDC-Net, a multi-path detection calibration network, to address the data distribution discrepancy between object proposals and refined bounding-boxes. Built on Faster R-CNN, PDC-Net utilizes a multi-path detection head to calibrate detection results and improve accuracy.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2021)
Article
Computer Science, Theory & Methods
Long Liu
Summary: This study conducts in-depth analysis of high-difficulty action recognition technology in basketball, and develops a big data motion target detection system based on deep convolutional neural network. It effectively enhances the training effect of sports through the recognition of difficult movements.
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE
(2021)
Article
Computer Science, Information Systems
Xiaoyu Chen, Hongliang Li, Qingbo Wu, Fanman Meng, Heqian Qiu
Summary: In this study, we propose Bal-(RCNN)-C-2 for high-quality recurrent object detection, with two new components that induce balanced optimization and result in significant improvement over existing solutions, reaching better performance than several state-of-the-art methods on evaluation datasets like PASCAL VOC and MSCOCO.
IEEE TRANSACTIONS ON MULTIMEDIA
(2022)
Article
Computer Science, Artificial Intelligence
Yong Xu, Chaoda Zheng, Ruotao Xu, Yuhui Quan, Haibin Ling
Summary: In recent years, multi-view learning has become a promising approach for 3D shape recognition by identifying shapes based on 2D views from different angles. This paper proposes a correspondence-aware representation (CAR) module that finds potential intra-view and cross-view correspondences through kNN search in semantic space and aggregates shape features via learned transforms. Incorporating the CAR module into a ResNet-18 backbone, an effective deep model called CAR-Net is introduced for 3D shape classification and retrieval, demonstrating the effectiveness and excellent performance of the CAR module.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2021)
Article
Computer Science, Information Systems
Joo Chan Lee, Yongwoo Kim, Sungtae Moon, Jong Hwan Ko
Summary: This study proposes a reconfigurable DNN architecture for real-time object detection, which can configure the optimal split point according to the edge-cloud CI environment. The proposed architecture allows for DNN model splitting through a feature reconstruction network and asymmetric scaling. Performance evaluation using YOLOv5 as the baseline showed significant improvement in inference speed compared to edge-only and cloud-only inference.
IEEE INTERNET OF THINGS JOURNAL
(2022)
Article
Psychology, Multidisciplinary
Fazal Wahab, Inam Ullah, Anwar Shah, Rehan Ali Khan, Ahyoung Choi, Muhammad Shahid Anwar
Summary: Computer vision and human-computer interaction are crucial in various technological fields. This study focuses on real-time object detection and recognition systems using the SSD algorithm and deep learning techniques to achieve high accuracy and efficiency.
FRONTIERS IN PSYCHOLOGY
(2022)
Article
Computer Science, Artificial Intelligence
Yafeng Zhang, Junyang Yu, Yuanyuan Wang, Shuang Tang, Han Li, Zhiyi Xin, Chaoyi Wang, Ziming Zhao
Summary: The ability of modern detectors to detect small targets is still a challenge compared to detecting medium and large targets in object detection. The proposed DLGADet method combines hierarchical attention mechanisms with deformable multi-scale feature fusion to improve recognition and detection performance. It introduces a combination of multi-scale separable detection and multi-scale feature fusion to obtain richer contextual information, addresses object deformations through a deformation feature extraction module, and incorporates a HAM combining global and local attention mechanisms to extract discriminative features. Extensive experiments on three datasets demonstrate the effectiveness of the proposed methods.
IET IMAGE PROCESSING
(2023)
Article
Computer Science, Theory & Methods
Yousung Yang, Seongsoo Lee, Joohyung Lee
Summary: This paper presents the design and implementation of a video analytics based real-time intelligent crossing detection system (RICDS) for smart cities. The system utilizes an adaptive queue management-based object tracking scheme to enhance object tracking on edge devices with limited computational resources. The system also introduces a real-world-real-time tracking scheme to predict the future positions of multiple objects and assign unique IDs to them. Experimental results show that the proposed tracking scheme achieves a significant latency reduction while maintaining similar multi object tracking accuracy compared to benchmark schemes.
FUTURE GENERATION COMPUTER SYSTEMS-THE INTERNATIONAL JOURNAL OF ESCIENCE
(2022)
Article
Computer Science, Information Systems
Sudan Jha, Changho Seo, Eunmok Yang, Gyanendra Prasad Joshi
Summary: This paper introduces a system that enables real-time video surveillance in low-end edge computing environments by combining object detection tracking algorithm. The study proposes N-YOLO, a method that divides images into fixed-size sub-images and merges detection results using correlation-based tracking algorithm to significantly reduce computation for object detection and tracking. Additionally, a system is proposed to guarantee real-time performance in various edge computing environments by adaptively controlling the cycle of object detection and tracking.
MULTIMEDIA TOOLS AND APPLICATIONS
(2021)
Article
Chemistry, Analytical
Qinghua Sheng, Haixiang Sheng, Peng Gao, Zhu Li, Haibing Yin
Summary: This study proposes a real-time detection scheme for cook assistant overalls based on the Hi3559A embedded processor, which improves the network reasoning speed on embedded devices through optimizing the network model and parallel processing technology. This allows for accurate image recognition and effective application in the scene of identifying kitchen overalls.
Article
Computer Science, Artificial Intelligence
Kan Huang, Chunwei Tian, Zhijing Xu, Nannan Li, Jerry Chun-Wei Lin
Summary: This paper proposes a Motion Context guided Edge-preserving network (MCE-Net) model for video salient object detection. MCE-Net can generate temporally consistent salient edges and refine the salient object regions completely and uniformly. Experimental results demonstrate the superior performance of the proposed method on five widely-used benchmarks.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Biochemistry & Molecular Biology
Jing Ning, Zhou Li, Xingjiang Zhang, Junlong Wang, Dandan Chen, Qiong Liu, Yi Sun
Summary: This study developed unbiased machine learning tools to quantify the behavior of fruit flies and discovered that fruit flies can extract higher-order features from conspecifics during courtship and select appropriate actions. The findings lay the foundation for understanding the mechanism of conspecific recognition in fruit flies.