Article
Computer Science, Artificial Intelligence
Tongqing Zhou, Zhiping Cai, Fang Liu, Jinshu Su
Summary: Propose a novel photo selection framework with adaptive aesthetic awareness for crowdsensing, which actively learns contextual knowledge for dynamically tailoring the aesthetic predictor. Extensive experiments demonstrate the performance superiority of CrowdPicker over baselines and sampling strategies.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2023)
Article
Robotics
Krzysztof Lis, Sina Honari, Pascal Fua, Mathieu Salzmann
Summary: This study combines road obstacle detection techniques with perspective information to address the issue of diminishing apparent size of obstacles as their distance from the vehicle increases. The results demonstrate that the combination of these two strategies significantly improves obstacle detection performance and outperforms existing methods in terms of instance-level obstacle detection.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2023)
Article
Robotics
Chuyan Zhang, Yun Gu, Jie Yang, Guang-Zhong Yang
Summary: The proposed diversity-aware learning framework predicts the optimal focus position based on a single image, utilizing a two-point representation of distance for label distribution learning and an intra-class discrepancy penalty term to reduce pathology slide variation. Experiments show promising results in accuracy, real-time performance, and generalization, outperforming previous no-reference approaches by 39%.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2021)
Article
Computer Science, Artificial Intelligence
Dongsheng Ruan, Yu Shi, Jun Wen, Nenggan Zheng, Min Zheng
Summary: This study introduces a novel SAC block for learning spatially-aware contexts to improve long-range dependency modeling, with extensive experiments showing significant performance improvements in various computer vision tasks.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2021)
Article
Computer Science, Artificial Intelligence
Sean Bin Yang, Chenjuan Guo, Bin Yang
Summary: Ranking paths in transportation services is an important functionality, and this study proposes a regression modeling approach to assign ranking scores to paths based on historical trajectories. The study introduces effective training data enrichment and a multi-task learning framework to improve the ranking estimation. Empirical studies validate the effectiveness and practicality of the proposed framework.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2022)
Article
Robotics
Francois Robinet, Yussef Akl, Kaleem Ullah, Farzad Nozarian, Christian Mueller, Raphael Frank
Summary: Identifying traversable space is a crucial problem in autonomous robot navigation, and recent research has focused on unsupervised and semi-supervised approaches to reduce annotation costs. This study proposes a practical and minimally-supervised method for monocular road segmentation, utilizing task-specific feature extraction and pseudo-labeling. The results show that even minimal labeling efforts can greatly improve the performance, demonstrating the pragmatic approach to labeling.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2022)
Article
Computer Science, Information Systems
Zhou Fang, Jiaxin Qi, Lubin Fan, Jianqiang Huang, Ying Jin, Tianren Yang
Summary: Existing deep-learning tools for road network generation have limitations in flat urban areas. This paper proposes a new method that combines geometric configurations and topographic features to automate street network generation in both flat and hilly urban areas. The improved model shows more realistic predictions and is more effective in generating streets in hilly areas. The geo-extractor module provides insights in recognizing and considering topographic information.
INTERNATIONAL JOURNAL OF GEOGRAPHICAL INFORMATION SCIENCE
(2022)
Article
Automation & Control Systems
Mengya Xu, Mobarakol Islam, Ben Glocker, Hongliang Ren
Summary: Curriculum learning and self-paced learning are effective training strategies in robotic vision. Most existing works focus on designing curricula based on difficulty levels, but the approach of smoothing labels for learning control is unexplored. In this work, a paced curriculum by label smoothing (P-CBLS) is proposed for classification and semantic segmentation tasks. The proposed method improves prediction accuracy and robustness.
IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING
(2023)
Article
Computer Science, Information Systems
Eugene Tan Boon Hong, Yung-Wey Chong, Tat-Chee Wan, Kok-Lim Alvin Yau
Summary: This paper proposes a new visual dialog dataset called DS-Dialog to address the challenges in obtaining sufficient context and overcoming visual semantic limitations faced by the existing dataset. DS-Dialog enhances the current dataset by adding new context-aware relevant history to provide more visual semantic context for each image. The proposed DS-Dialog model achieves better performance compared to previous models, demonstrating the importance of relevant semantic historical context in enhancing the visual semantic relationship between textual and visual representations.
CMC-COMPUTERS MATERIALS & CONTINUA
(2023)
Article
Engineering, Civil
Amr Abdelraouf, Mohamed Abdel-Aty, Yina Wu
Summary: This paper proposes a novel vision-based method to detect rain and road surface conditions using roadside traffic cameras. By utilizing vision transformers for image-based classification and leveraging the geographical distribution of roadside cameras and a spatial self-attention network, the detection model's accuracy is enhanced.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2022)
Article
Automation & Control Systems
Chen Zhu, Jianyu Yang, Zhanpeng Shao, Chunping Liu
Summary: The paper introduces a new method for hand gesture recognition using depth maps and 3D shape context descriptors. Experimental results show that the proposed method is robust to noise, articulated variations, and rigid transformations, outperforming current state-of-the-art methods in accuracy and efficiency.
IEEE-CAA JOURNAL OF AUTOMATICA SINICA
(2021)
Article
Chemistry, Analytical
Weipeng Shi, Wenhu Qin, Zhonghua Yun, Peng Ping, Kaiyang Wu, Yuke Qu
Summary: This study introduces an end-to-end semantic segmentation model for aerial images based on HRNET, addressing two challenges in RSIs semantic segmentation with the incorporation of CRAM and CCFM modules. Experimental results show that the model improves accuracy and outperforms some commonly used CNN architectures.
Article
Computer Science, Information Systems
Lu Yang, Qing Song, Zhihui Wang, Zhiwei Liu, Songcen Xu, Zhihao Li
Summary: This study proposes a statistical method based on the output probability map to estimate the quality of network output for human parsing. It introduces a Quality-Aware Module (QAM) and a Quality-Aware Network (QANet) to improve the quality of human parsing results. The method achieves the best performance on multiple benchmarks and has the potential for application in other tasks.
IEEE TRANSACTIONS ON MULTIMEDIA
(2023)
Article
Computer Science, Artificial Intelligence
Haobo Yuan, Teng Chen, Wei Sui, Jiafeng Xie, Lefei Zhang, Yuan Li, Qian Zhang
Summary: This paper proposes a novel deep neural network called RPANet for 3D sensing from monocular image sequences based on planar parallax. RPANet takes a pair of images aligned by the homography of the road plane as input and outputs a height-to-depth ratio map for 3D reconstruction. The map can be combined with the road plane as a reference to estimate the 3D structure by warping the consecutive frames. The effectiveness of the method is verified through comprehensive experiments on the Waymo Open Dataset.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2023)
Article
Computer Science, Artificial Intelligence
Jialu Zhang, Jianfeng Ren, Qian Zhang, Jiang Liu, Xudong Jiang
Summary: In this paper, a multi-branch deep neural network is proposed for multi-label image classification. It effectively utilizes label-related semantic information, background context, and spatial semantic information to better detect target objects. Experimental results show that the proposed method outperforms the state-of-the-art methods for multi-label image classification.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2023)
Article
Geochemistry & Geophysics
Mingwei Zhang, Qiang Li, Yuan Yuan, Qi Wang
Summary: This article proposes a building change detection method based on deep learning, which effectively addresses the challenges of temporal-spatial correlation and discrimination in the neighborhood of the edge by introducing a selective attention module and a contrastive learning method. The experimental results show that the proposed method achieves competitive performance in terms of objective metrics and visual comparisons.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2023)
Article
Engineering, Electrical & Electronic
Haoxuan Ding, Junyu Gao, Yuan Yuan, Qi Wang
Summary: Scene Text Detection (STD) has been successfully applied in various fields. One important application is License Plate Detection (LPD). License Plate (LP) serves as a unique identifier for vehicles, facilitating intelligent transportation in areas such as traffic enforcement and dispatching. However, similar scene texts often lead to misjudgment by LP detectors. To address this issue, more discriminative features are required. In this study, we propose a Self-Constrained Contrastive Aggregation (SCCA) method to aggregate features in the latent space and improve the feature expression of the backbone. Experimental results demonstrate that SCCA significantly improves the baseline performance and outperforms recent LP detectors, achieving a 99.7 F1-score and AP on the UFPR-ALPR dataset. Additionally, a comparison between self-constrained contrastive learning and vanilla contrastive learning confirms the superiority of SCCA and the reasonability of our assumptions.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2023)
Article
Geochemistry & Geophysics
Cong Zhang, Kin-Man Lam, Qi Wang
Summary: In this article, a novel coarse-to-fine framework (CoF-Net) is proposed for object detection in remote-sensing imagery, which aims to improve the performance of existing object detectors by progressively enhancing feature representation and selecting stronger training samples. CoF-Net smoothly refines the original coarse features into multispectral nonlocal fine features with discriminative spatial-spectral details and semantic relations, and dynamically considers samples from coarse to fine during training by introducing geometric and classification constraints. Comprehensive experiments demonstrate the effectiveness and superiority of the proposed method.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Computer Science, Artificial Intelligence
Qi Wang, Xiaocheng Lu, Cong Zhang, Yuan Yuan, Xuelong Li
Summary: This paper constructs a large-scale video-based license plate dataset named LSV-LP, and proposes a new framework called MFLPR-Net to improve the performance of license plate detection and recognition systems.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Geography, Physical
Wei Huang, Yilei Shi, Zhitong Xiong, Qi Wang, Xiao Xiang Zhu
Summary: RS image scene classification has gained attention for its applications. Conventional supervised approaches require labeled data, but with more RS images available, utilizing unlabeled data becomes urgent. This paper proposes a SSDA method called BSCA for RS cross-domain scene classification, using unsupervised and supervised alignment strategies to reduce domain shift.
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING
(2023)
Article
Geochemistry & Geophysics
Jingtao Hu, Junyu Gao, Yuan Yuan, Jocelyn Chanussot, Qi Wang
Summary: The study proposes a location-guided network (LGNet) to improve connectivity performance in road extraction. By adding an auxiliary road location prediction (RLP) task, LGNet obtains global road connectivity information and enhances road segmentation performance. The features are guided by the global location context using a location-guided decoder (LG-Decoder) to capture the connectivity of each road segment.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Engineering, Civil
Haoxuan Ding, Junyu Gao, Yuan Yuan, Qi Wang
Summary: License Plate (LP) is crucial for intelligent transportation, and a contrastive learning method is proposed to improve the detection accuracy. The Contrastive License Plate Detector (CLPD) employs a contrastive triad to decouple the foregrounds and backgrounds, and a contrastive learning branch is introduced to enhance the feature expression ability and extract discriminative features. The CLPD achieves significant improvements compared to baselines and other license plate detectors on multiple datasets.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Geochemistry & Geophysics
Cong Zhang, Jingran Su, Yakun Ju, Kin-Man Lam, Qi Wang
Summary: A novel efficient inductive vision Transformer framework is proposed for oriented object detection in remote sensing imagery. The framework fully explores spatial redundancy and utilizes an adaptive multigrained routing mechanism to reduce computational cost without compromising accuracy. It also incorporates a compact dual-path encoding architecture and an angle tokenization technique to enhance inductive bias and promote the encoding and learning of directional knowledge. Comprehensive experiments demonstrate the effectiveness and superiority of the proposed framework for oriented object detection in remote sensing images.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Geochemistry & Geophysics
Yanfeng Liu, Yuan Yuan, Qi Wang
Summary: In this paper, a hybrid modeling approach called the uncertainty-aware graph reasoning with global collaborative learning (UG2L) framework is proposed to address the accurate detection problem of salient objects in optical remote sensing images (RSIs) with complex edges and irregular topology. The proposed method models the intricate relations among RSI patches using graph representations. Additionally, a global context block with a linear attention mechanism is introduced to explore the multiscale and global context collaboratively, and an uncertainty-aware loss is designed to enhance the model's reliability for better saliency prediction.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2023)
Article
Geochemistry & Geophysics
Yanfeng Liu, Zhitong Xiong, Yuan Yuan, Qi Wang
Summary: Existing remote sensing image salient object detection methods often focus on pixel-level supervision and overlook image-level scene information. In this study, we annotate image-level scene labels for three datasets and propose a scene-guided dual-branch network (SDNet) that performs cross-task knowledge distillation to improve saliency detection accuracy.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Geochemistry & Geophysics
Yuyu Jia, Junyu Gao, Wei Huang, Yuan Yuan, Qi Wang
Summary: This paper proposes a multiview integration network (HSL-MINet) to tackle the few-shot learning problem in remote sensing scene classification. The network enhances the model's generalization ability and discriminative power of the decision boundary through a multiview integration module and a hard sample learning module. Extensive experiments on multiple datasets demonstrate that HSL-MINet outperforms previous state-of-the-art methods.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Geochemistry & Geophysics
Yanfeng Liu, Zhitong Xiong, Yuan Yuan, Qi Wang
Summary: This study proposes a universal super-resolution assisted learning (SRAL) framework to improve the performance and efficiency of existing remote sensing salient object detectors. The framework includes a transposed saliency detection decoder (TSDD), an auxiliary SR decoder (ASRD), and a task-fusion guidance module (TFGM). Experimental results on three datasets demonstrate that SRAL outperforms more than 20 algorithms and can be applied to existing networks.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Geochemistry & Geophysics
Yuyu Jia, Junyu Gao, Wei Huang, Yuan Yuan, Qi Wang
Summary: This article proposes a holistic mutual representation enhancement method for few-shot segmentation, addressing the issues of intra-class variations and background interference. Extensive experiments demonstrate the superiority of the proposed method and a corresponding benchmark dataset is created.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Engineering, Electrical & Electronic
Cong Zhang, Tianshan Liu, Jun Xiao, Kin-Man Lam, Qi Wang
Summary: This article proposes a novel pretraining paradigm specifically for RS object detection, which significantly improves detection performance and outperforms traditional classification pretraining methods. The method generates pseudo bounding boxes on a reconstructed RS classification-style dataset and integrates them with accurate class labels as location- and category-related supervisions for pretraining the RS detectors.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Article
Geochemistry & Geophysics
Qiang Li, Yuan Yuan, Qi Wang
Summary: This paper proposes a multiscale factor joint learning method for hyperspectral image super-resolution, which effectively explores the interdependence among different scale factors and optimizes the representation of spatial and spectral contents through information interaction and feedback correction, achieving superior performance.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Computer Science, Artificial Intelligence
Rui Lv, Dingheng Wang, Jiangbin Zheng, Zhao-Xu Yang
Summary: In this paper, the authors investigate tensor decomposition for neural network compression. They analyze the convergence and precision of tensor mapping theory, validate the rationality of tensor mapping and its superiority over traditional tensor approximation based on the Lottery Ticket Hypothesis. They propose an efficient method called 3D-KCPNet to compress 3D convolutional neural networks using the Kronecker canonical polyadic (KCP) tensor decomposition. Experimental results show that 3D-KCPNet achieves higher accuracy compared to the original baseline model and the corresponding tensor approximation model.
Article
Computer Science, Artificial Intelligence
Xiangkun He, Zhongxu Hu, Haohan Yang, Chen Lv
Summary: In this paper, a novel constrained multi-objective reinforcement learning algorithm is proposed for personalized end-to-end robotic control with continuous actions. The approach trains a single model using constraint design and a comprehensive index to achieve optimal policies based on user-specified preferences.
Article
Computer Science, Artificial Intelligence
Zhijian Zhuo, Bilian Chen, Shenbao Yu, Langcai Cao
Summary: In this paper, a novel method called Expansion with Contraction Method for Overlapping Community Detection (ECOCD) is proposed, which utilizes non-negative matrix factorization to obtain disjoint communities and applies expansion and contraction processes to adjust the degree of overlap. ECOCD is applicable to various networks with different properties and achieves high-quality overlapping community detection.
Article
Computer Science, Artificial Intelligence
Yizhe Zhu, Chunhui Zhang, Jialin Gao, Xin Sun, Zihan Rui, Xi Zhou
Summary: In this work, the authors propose a Contrastive Spatio-Temporal Distilling (CSTD) approach to improve the detection of high-compressed deepfake videos. The approach leverages spatial-frequency cues and temporal-contrastive alignment to fully exploit spatiotemporal inconsistency information.
Review
Computer Science, Artificial Intelligence
Laijin Meng, Xinghao Jiang, Tanfeng Sun
Summary: This paper provides a review of coverless steganographic algorithms, including the development process, known contributions, and general issues in image and video algorithms. It also discusses the security of coverless steganography from theoretical analysis to actual investigation for the first time.
Article
Computer Science, Artificial Intelligence
Yajie Bao, Tianwei Xing, Xun Chen
Summary: Visual question answering requires processing multi-modal information and effective reasoning. Neural-symbolic learning is a promising method, but current approaches lack uncertainty handling and can only provide a single answer. To address this, we propose a confidence based neural-symbolic approach that evaluates NN inferences and conducts reasoning based on confidence.
Article
Computer Science, Artificial Intelligence
Anh H. Vo, Bao T. Nguyen
Summary: Interior style classification is an interesting problem with potential applications in both commercial and academic domains. This project proposes a method named ISC-DeIT, which combines data-efficient image transformer architectures and knowledge distillation, to address the interior style classification problem. Experimental results demonstrate a significant improvement in predictive accuracy compared to other state-of-the-art methods.
Article
Computer Science, Artificial Intelligence
Shashank Kotyan, Danilo Vasconcellos Vargas
Summary: This article introduces a novel augmentation technique called Dynamic Scanning Augmentation to improve the accuracy and robustness of Vision Transformer (ViT). The technique leverages dynamic input sequences to adaptively focus on different patches, resulting in significant changes in ViT's attention mechanism. Experimental results demonstrate that Dynamic Scanning Augmentation outperforms ViT in terms of both robustness to adversarial attacks and accuracy against natural images.
Article
Computer Science, Artificial Intelligence
Hiba Alqasir, Damien Muselet, Christophe Ducottet
Summary: The article proposes a solution to improve the learning process of a classification network by providing shape priors, reducing the need for annotated data. The solution is tested on cross-domain digit classification tasks and a video surveillance application.
Article
Computer Science, Artificial Intelligence
Dexiu Ma, Mei Liu, Mingsheng Shang
Summary: This paper proposes a method using neural dynamics solvers to solve infinity-norm optimization problems. Two improved solvers are constructed and their effectiveness and superiority are demonstrated through theoretical analysis and simulation experiments.
Article
Computer Science, Artificial Intelligence
Francesco Gregoretti, Giovanni Pezzulo, Domenico Maisto
Summary: Active Inference is a computational framework that uses probabilistic inference and variational free energy minimization to describe perception, planning, and action. cpp-AIF is a header-only C++ library that provides a powerful tool for implementing Active Inference for Partially Observable Markov Decision Processes through multi-core computing. It is cross-platform and improves performance, memory management, and usability compared to existing software.
Article
Computer Science, Artificial Intelligence
Zelin Ying, Dawei Cheng, Cen Chen, Xiang Li, Peng Zhu, Yifeng Luo, Yuqi Liang
Summary: This paper proposes a novel stock market trends prediction framework called SMART, which includes a self-supervised stock technical data sequence embedding model S3E. By training with multiple self-supervised auxiliary tasks, the model encodes stock technical data sequences into embeddings and uses the learned sequence embeddings for predicting stock market trends. Extensive experiments on China A-Shares market and NASDAQ market prove the high effectiveness of our model in stock market trends prediction, and its effectiveness is further validated in real-world applications in a leading financial service provider in China.
Article
Computer Science, Artificial Intelligence
Hao Li, Hao Jiang, Dongsheng Ye, Qiang Wang, Liang Du, Yuanyuan Zeng, Liu Yuan, Yingxue Wang, C. Chen
Summary: DHGAT1, a dynamic hyperbolic graph attention network, utilizes hyperbolic metric properties to embed dynamic graphs. It employs a spatiotemporal self-attention mechanism and weighted node representations, resulting in excellent performance in link prediction tasks.
Article
Computer Science, Artificial Intelligence
Jiehui Huang, Zhenchao Tang, Xuedong He, Jun Zhou, Defeng Zhou, Calvin Yu-Chian Chen
Summary: This study proposes a progressive learning multi-scale feature blending model for image deraining tasks. The model utilizes detail dilation and texture extraction to improve the restoration of rainy images. Experimental results show that the model achieves near state-of-the-art performance in rain removal tasks and exhibits better rain removal realism.
Article
Computer Science, Artificial Intelligence
Lizhi Liu, Zilin Gao, Yinhe Wang, Yongfu Li
Summary: This paper proposes a novel discrete-time interconnected model for depicting complex dynamical networks. The model consists of nodes and edges subsystems, which consider the dynamic characteristic of both nodes and edges. By designing control strategies and coupling modes, the stabilization and synchronization of the network are achieved. Simulation results demonstrate the effectiveness of the proposed methods.