Article
Geography, Physical
K. R. Akshatha, A. K. Karunakar, B. Satish Shenoy, K. Phani Pavan, Chinmay V. Dhareshwar, Dennis George Johnson
Summary: Intelligent UAV video analysis has gained attention for its potential in computer vision applications. In order to address the challenge of small object detection, a Manipal-UAV person detection dataset was created, consisting of images captured from UAVs in varying conditions. The dataset provides a benchmark for evaluating state-of-the-art object detection algorithms on small person objects in aerial view scenarios. The dataset is publicly available for researchers to advance UAV and small object detection research.
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING
(2023)
Article
Engineering, Electrical & Electronic
Tung Minh Tran, Tu N. Vu, Tam V. Nguyen, Khang Nguyen
Summary: There is a lack of drone-based datasets for detecting abnormal events in urban traffic environments, especially roundabouts. To address this issue, researchers have developed a large-scale drone dataset consisting of 51 videos recorded at roundabouts in Vietnam, covering various abnormal event types. This dataset is used to evaluate the performance of state-of-the-art algorithms in drone-based video surveillance for anomaly detection.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
(2023)
Article
Computer Science, Artificial Intelligence
Abdelrahman Abdallah, Alexander Berendeyev, Islam Nuradin, Daniyar Nurseitov
Summary: In this paper, we present the TNCR dataset, which consists of tables with varying image quality collected from open access websites. The dataset can be used for table detection and classification. We have implemented state-of-the-art deep learning-based methods for table detection and achieved the best performance with the Deformable DERT model with Resnet-50 Backbone Network on the TNCR dataset.
Article
Computer Science, Information Systems
Woo-Han Yun, Taewoo Kim, Jaeyeon Lee, Jaehong Kim, Junmo Kim
Summary: Training an object instance detector with a cut-and-paste method may lead to a conventional domain shift problem, which can be divided into two sub-domain gaps for foreground and background. A new advanced cut-and-paste method is introduced to balance the unbalanced domain gaps by diversifying the foreground with GAN-generated seed images and simplifying the background using image processing techniques, effectively improving the accuracy of object instance detection in cluttered indoor environments with limited seed images while also enhancing the detection accuracy of state-of-the-art domain adaptation methods.
Article
Chemistry, Analytical
Jing Xie, Erik Stensrud, Torbjorn Skramstad
Summary: The system proposed includes two stages to automatically process maritime ship inspection videos and predict suspicious areas where cracks may exist. It has demonstrated the feasibility of applying deep neural network-based computer vision technologies to automating remote ship inspection, achieving reasonable performance in tracking and analyzing ship inspection videos.
Article
Computer Science, Information Systems
Mengyao Zhang, Xianwei Rong, Xiaoyan Yu
Summary: This article presents a lightweight convolutional neural network (Light-SDNet) for ship detection under different weather conditions. By introducing improved convolution operations and parallel attention, and using a hybrid training strategy with generated synthetic images to enhance model adaptability, the ship detection results under severe weather conditions are improved.
Article
Engineering, Electrical & Electronic
Shuang Yang, Wentao An, Shibao Li, Gengli Wei, Bin Zou
Summary: This paper introduces an improved fully convolutional one-stage object detector (Improved-FCOS) to address ship detection in synthetic aperture radar (SAR) images. The method proposes a multilevel feature attention mechanism, a feature refinement and reuse module, and a head improvement module to enhance the accuracy and robustness of ship detection. Experimental results demonstrate that Improved-FCOS achieves the best detection performance on multiple datasets.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
(2022)
Article
Construction & Building Technology
An Xuehui, Zhou Li, Liu Zuguang, Wang Chengzhi, Li Pengfei, Li Zhiwei
Summary: The study introduced the MOCS image dataset for detecting moving objects in construction sites, featuring precise annotations and statistical analysis for accurate object localization. All detectors trained on the dataset were able to detect objects at construction sites precisely and robustly.
AUTOMATION IN CONSTRUCTION
(2021)
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
Chemistry, Analytical
Arkadiusz Bozko, Leszek Ambroziak
Summary: Robotic systems like UAVs and USVs are increasingly used for objects and events detection tasks. Deep learning, particularly CNN, is incorporated in these solutions due to the low cost of autonomous vehicles, vision sensors, and portable computers. Insufficient training datasets can decrease the effectiveness of the systems, and dataset augmentation is a solution to enhance the datasets. This research evaluates the influence of training image augmentation on the important parameters for neural networks in object detection.
Article
Environmental Sciences
Linhao Li, Zhiqiang Zhou, Bo Wang, Lingjuan Miao, Zhe An, Xiaowu Xiao
Summary: This paper proposes a domain adaptive ship detection method that minimizes domain discrepancies through image-level and instance-level adaption, enhancing the accuracy and generalization ability of ship detection.
Article
Environmental Sciences
Li Shen, Yao Lu, Hao Chen, Hao Wei, Donghai Xie, Jiabao Yue, Rui Chen, Shouye Lv, Bitao Jiang
Summary: This study introduces a building-change-detection dataset named S2Looking, which consists of large-scale side-looking satellite images and tens of thousands of annotated change instances, for training deep-learning algorithms. The dataset offers larger viewing angles, illumination variances, and complexity of rural images compared to existing datasets, and preliminary tests suggest higher level of challenges for deep-learning algorithms.
Article
Computer Science, Artificial Intelligence
Yuni Zeng, Qianwen Duan, Xiangru Chen, Dezhong Peng, Yao Mao, Ke Yang
Summary: Unmanned aerial vehicles (UAVs) play a significant role in enhancing convenience and intelligence in life, but their large-scale use also raises security concerns. Advanced object detection models driven by data are crucial for UAV detection tasks. The UAVData dataset, containing well-labeled images of balloons and 6 types of UAVs, serves as a valuable resource for researchers to advance in both UAV detection and machine learning applications.
Article
Geochemistry & Geophysics
Feng Zhang, Xueying Wang, Shilin Zhou, Yingqian Wang, Yi Hou
Summary: In this article, we propose a center-head point extraction-based detector (CHPDet) for arbitrary-oriented ship detection in remote sensing images. Experimental results demonstrate that our CHPDet achieves state-of-the-art performance on a new dataset FGSD2021 and two other commonly used datasets.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Computer Science, Interdisciplinary Applications
Xiaopei Cai, Xueyang Tang, Shuo Pan, Yi Wang, Hai Yan, Yuheng Ren, Ning Chen, Yue Hou
Summary: This article introduces an intelligent recognition method for defects in concrete slabs of high-speed railway (HSR) based on a few-shot learning model. By training an artificial intelligence model with limited data, it can accurately identify different service conditions of HSR. The experimental results show satisfactory accuracy, low parameter quantity, and testing time.
COMPUTER-AIDED CIVIL AND INFRASTRUCTURE ENGINEERING
(2023)
Article
Engineering, Electrical & Electronic
Jianyu Chen, Zhongyuan Wang, Kangli Zeng, Zheng He, Zixiang Xiong
Summary: In the task of video action recognition, we propose a lightweight video feature extraction strategy that performs convolution on the video cube from three orthogonal angles to learn appearance and motion features. Our approach achieves better recognition performance with lower computational volume than 3D CNN.
IEEE SIGNAL PROCESSING LETTERS
(2022)
Article
Computer Science, Artificial Intelligence
Shanfa Ke, Zhongyuan Wang, Ruimin Hu, Xiaochen Wang
Summary: This paper proposes a single-channel multi-speaker speech separation method based on the similarity between the speaker feature center and the mixture feature in the deep embedding space. The method achieves separation of speaker's speech by extracting isolated speech segments and creating a speaker feature center. A residual-based deep embedding network is introduced for faster speed and better feature extraction. Experimental results show that the proposed method outperforms competing algorithms in Signal-to-Distortion Ratio.
NEURAL PROCESSING LETTERS
(2023)
Article
Computer Science, Information Systems
Baojin Huang, Zhongyuan Wang, Guangcheng Wang, Zhen Han, Kui Jiang
Summary: During the COVID-19 pandemic, wearing masks has become ubiquitous, posing a challenge for traditional face recognition algorithms. This article proposes a local eyebrow feature attention network for recognizing masked faces, which utilizes the identification potential of eyebrow features due to the limited visible information on masked faces. The network includes feature extraction, eyebrow region pooling, and feature fusion, achieving superior performance compared to state-of-the-art methods on benchmark datasets.
ACM TRANSACTIONS ON MULTIMEDIA COMPUTING COMMUNICATIONS AND APPLICATIONS
(2023)
Article
Computer Science, Information Systems
Zhihao Shen, Kang Yang, Xi Zhao, Jianhua Zou, Wan Du
Summary: This paper aims to predict a set of apps a user will open on her mobile device in the next time slot. Such information is essential for smartphone operations to improve user experience. By using the deep reinforcement learning framework DeepAPP, which learns from historical app usage data and adapts to time-varying behavior, the authors address the challenges of accurately capturing complex contextual environment and predicting a set of apps.
IEEE TRANSACTIONS ON MOBILE COMPUTING
(2023)
Article
Computer Science, Artificial Intelligence
Kunpeng Liu, Dongjie Wang, Wan Du, Dapeng Oliver Wu, Yanjie Fu
Summary: In this paper, a single-agent Monte Carlo-based reinforced feature selection method is proposed, along with two efficiency improvement strategies: early stopping strategy and reward-level interactive strategy. The proposed method aims to find the optimal feature subset for a given machine learning task by traversing the feature set and selecting features one by one. Additionally, the early stopping strategy and reward-level interactive strategy are introduced to enhance the training efficiency.
KNOWLEDGE AND INFORMATION SYSTEMS
(2023)
Article
Engineering, Electrical & Electronic
Yuanzhi Wang, Tao Lu, Yanduo Zhang, Zhongyuan Wang, Junjun Jiang, Zixiang Xiong
Summary: In this paper, the authors propose a method called FaceFormer, which combines global features from Transformers and local features from CNNs to restore high-quality face images.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2023)
Article
Computer Science, Artificial Intelligence
Kui Jiang, Zhongyuan Wang, Peng Yi, Chen Chen, Guangcheng Wang, Zhen Han, Junjun Jiang, Zixiang Xiong
Summary: This study focuses on addressing the problem of generating rain-free images under complex rain conditions using deep learning models. By designing a multi-level pyramid structure, non-local fusion module, attention fusion module, and residual learning branch to handle different challenges, the results demonstrate that our method achieves superior performance in generating rain-free images.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Xin Tian, Kun Li, Wei Zhang, Zhongyuan Wang, Jiayi Ma
Summary: In this study, we propose an interpretable model-driven deep network called HMPNet for hyperspectral, multispectral, and panchromatic image fusion. The model utilizes a deep prior to describe the relationship between high-resolution hyperspectral and panchromatic images. By solving the optimization problem using the proximal gradient descent algorithm and unrolling the iterative steps into network modules, we obtain the HMPNet. The HMPNet simplifies parameter selection, achieves a balance between spatial and spectral qualities, and improves generalization capability with modules that have explainable physical meanings.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Baojin Huang, Zhongyuan Wang, Kui Jiang, Qin Zou, Xin Tian, Tao Lu, Zhen Han
Summary: This study proposes a joint segmentation and identification feature learning framework for occlusion face recognition and solves the limitations of existing algorithms that focus on visible facial components. The method includes an occlusion prediction module and a channel refinement network to address this issue. The study also builds large-scale synthetic and real-world occlusion face datasets for experiments, and the results show significant improvements in face verification and face identification tasks.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Information Systems
Haiqiao Wu, Wan Du, Peng Gong, Dapeng Oliver Wu
Summary: In this article, a parallel Cross-rack Pipelining Update scheme (CPU) is proposed to achieve efficient update in distributed storage systems (DSSs). By dividing the update information into small units and transmitting them in parallel along multiple racks, CPU reduces update time and improves performance. The optimization of update path and slice size in CPU contributes to its effectiveness in reducing average update time.
IEEE TRANSACTIONS ON CLOUD COMPUTING
(2022)
Proceedings Paper
Computer Science, Information Systems
Xianzhong Ding, Wan Du
Summary: This paper presents a deep reinforcement learning-based irrigation system that improves irrigation efficiency by considering both current and future soil moisture loss. Through experiments, it is found that the system can save up to 9.52% of water.
2022 21ST ACM/IEEE INTERNATIONAL CONFERENCE ON INFORMATION PROCESSING IN SENSOR NETWORKS (IPSN 2022)
(2022)
Proceedings Paper
Computer Science, Information Systems
Xianzhong Ding, Wan Du
Summary: This research proposes an IoT-based irrigation system that utilizes advanced control algorithms and sensing technology to improve irrigation efficiency and conserve water resources.
2022 21ST ACM/IEEE INTERNATIONAL CONFERENCE ON INFORMATION PROCESSING IN SENSOR NETWORKS (IPSN 2022)
(2022)
Article
Computer Science, Artificial Intelligence
Tao Lu, Yuanzhi Wang, Yanduo Zhang, Junjun Jiang, Zhongyuan Wang, Zixiang Xiong
Summary: This article proposes a novel pre-prior guided approach that extracts facial prior information from high-resolution face images and embeds them into low-resolution ones, resulting in high-frequency information-rich low-resolution face images and improved face reconstruction performance.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Engineering, Electrical & Electronic
Xiya Rao, Tao Lu, Zhongyuan Wang, Yanduo Zhang
Summary: In this paper, a frequency guided network (FGNet) is proposed for few-shot semantic segmentation. The FGNet explicitly models the semantic information of different frequencies and guides the semantic alignment of the object. Experimental results show that FGNet achieves superior performance compared to state-of-the-art methods.
IEEE SIGNAL PROCESSING LETTERS
(2022)
Article
Computer Science, Information Systems
Zhenfeng Shao, Gui Cheng, Jiayi Ma, Zhongyuan Wang, Jiaming Wang, Deren Li
Summary: This article proposes a method for social distance monitoring using UAV, which utilizes a lightweight pedestrian detection network to detect pedestrians in real-time and calculate the social distance between them. Experimental results show that the proposed method outperforms traditional models on different datasets and enables real-time detection.
IEEE TRANSACTIONS ON MULTIMEDIA
(2022)