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, Artificial Intelligence
Dan-Sebastian Bacea, Florin Oniga
Summary: In this paper, we propose Mini-YOLOv4-tiny, an improved lightweight one-stage object detector based on YOLOv4-tiny. We achieve model compression by selectively cutting the width of the last convolutional layers while maintaining performance. Additionally, we propose several improvements that enhance the model's receptive field and achieve state-of-the-art results among lightweight object detectors.
IMAGE AND VISION COMPUTING
(2023)
Review
Computer Science, Information Systems
Ayoub Benali Amjoud, Mustapha Amrouch
Summary: This paper examines the evolution of object detection in the era of deep learning, reviews various state-of-the-art algorithms and their underlying concepts, and classifies them into anchor-based, anchor-free, and transformer-based detectors. The paper discusses the insights behind these algorithms and provides experimental analyses comparing quality metrics, speed/accuracy trade-offs, and training methodologies. Additionally, it compares major convolutional neural networks for object detection, highlights the strengths and limitations of each model, and summarizes the development of object detection methods under deep learning through simple graphical illustrations. Finally, the paper identifies future research directions.
Article
Chemistry, Analytical
Praneel Chand, Sunil Lal
Summary: This paper presents vision-based methods for detecting and classifying used electronics parts, utilizing shallow neural networks, support vector machines, and deep learning CNN models. Through a customized object detection algorithm in a multiple object workspace scenario, an overall accuracy of 98.1% was achieved.
Article
Geochemistry & Geophysics
Qi Ming, Lingjuan Miao, Zhiqiang Zhou, Yunpeng Dong
Summary: This article discusses the role of discriminative features in object detection and proposes a critical feature capturing network (CFC-Net) to improve detection accuracy by building powerful feature representation, refining preset anchors, and optimizing label assignment.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Review
Computer Science, Artificial Intelligence
Xuan-Thuy Vo, Kang-Hyun Jo
Summary: This paper provides a comprehensive review and analysis of anchor assignment and sampling methods in deep learning-based object detection. It identifies the strengths and weaknesses of each problem and discusses current issues and new research trends. The paper also offers a webpage for tracking the latest research progress.
Article
Geochemistry & Geophysics
Dehui Zhu, Bo Du, Liangpei Zhang
Summary: This article proposes a two-stream convolutional network-based target detector for hyperspectral images, extracting abundant spectral information and utilizing a hybrid sparse representation and classification strategy to select typical background pixels. A novel synthesis method is introduced to generate target samples, training the network with target and background samples to achieve superior performances in target detection.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2021)
Article
Environmental Sciences
Yuman Yuan, Hongyang Bai, Panfeng Wu, Hongwei Guo, Tianyu Deng, Weiwei Qin
Summary: A novel context sensing-YOLOv5 (CS-YOLOv5) is proposed for small and weak space object detection, which extracts local context information and enhances spatial information through the cross-layer context fusion module (CCFM) and adaptive weighting module (AWM). To address the issue of lost spatial information for small objects, a spatial information enhancement module (SIEM) is designed. Furthermore, a contrast mosaic data augmentation is proposed to enhance the generalization ability of CS-YOLOv5. Extensive experiments on self-built datasets prove the effectiveness of the method in space object detection.
Article
Engineering, Electrical & Electronic
Jiajia Liao, Yingchao Piao, Jinhe Su, Guorong Cai, Xingwang Huang, Long Chen, Zhaohong Huang, Yundong Wu
Summary: This article proposed an unsupervised cluster guided detection framework to address the issue of limited detection ability in scenes where objects are densely distributed in high-resolution aerial images. Experimental results show that the method outperforms existing baseline methods in two popular aerial image datasets.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
(2021)
Article
Geochemistry & Geophysics
Mridul Gupta, Jonathan Chan, Mitchell Krouss, Greg Furlich, Paul Martens, Moses W. Chan, Mary L. Comer, Edward J. Delp
Summary: This letter presents a method for infrared small target detection using convolutional neural networks (CNNs). The proposed method improves the probability of detection at low false detection rates.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Computer Science, Artificial Intelligence
Yang Liu, Peng Sun, Nickolas Wergeles, Yi Shang
Summary: This paper reviews deep learning methods for small object detection, discussing challenges, solutions, and techniques. Experimental results show that Faster R-CNN performs the best in detecting small objects.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Computer Science, Artificial Intelligence
Akshay Badola, Cherian Roy, Vineet Padmanabhan, Rajendra Prasad Lal
Summary: Interpretability of deep neural networks is difficult due to the lack of transparency in their decision-making process. This study proposes an algorithm to identify the most influential features in the final and penultimate layers of deep convolutional networks for image classification. The algorithm utilizes a decomposed softmax approach and achieves better performance compared to existing methods. The resulting low-dimensional embedding per class enhances interpretability and facilitates diagnosing misclassifications and pruning of the network.
NEURAL COMPUTING & APPLICATIONS
(2023)
Article
Computer Science, Information Systems
Chen Zhang, Zhengyu Xia, Joohee Kim
Summary: This paper proposes a video object detection network using event-aware ConvLSTM and object relation networks to improve detection performance under challenging conditions such as aspect ratio change, occlusion, or large motion. The proposed method enhances the ability to exploit temporal contextual information for video-based object detectors, achieving a mAP of 81.0% on the ImageNet VID dataset without post processing.
Article
Computer Science, Artificial Intelligence
Nuo Xu, Chunlei Huo, Xin Zhang, Chunhong Pan
Summary: This paper proposes a novel coarse-to-fine object detection method AHDet, which outperforms traditional approaches by using a two-step coarse-to-fine gaze process.
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
Geochemistry & Geophysics
Rui Zhao, Zhenwei Shi, Zhengxia Zou
Summary: An attention-based method that generates pixelwise semantic content segmentation masks for remote sensing images is proposed, achieving higher captioning accuracy compared to state-of-the-art methods. This method utilizes fine-grained, structured attention to exploit the structural characteristics of semantic contents in high-resolution remote sensing images.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Computer Science, Artificial Intelligence
Tianyang Shi, Zhengxia Zou, Zhenwei Shi, Yi Yuan
Summary: This paper proposes a new method to automatically create game characters based on a single face photo, utilizing a imitator network and gradient descent for optimization. Unlike previous methods, our approach produces fine-grained facial parameters with physical significance, allowing users to fine-tune their characters.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Engineering, Civil
Rusheng Zhang, Zhengxia Zou, Shengyin Shen, Henry X. Liu
Summary: This paper introduces a newly developed and deployed roadside cooperative perception system with an edge-cloud structure and multiple kinds of sensors. The performance of the system is analyzed using data collected from the field, and its potential in applications such as traffic volume monitoring and road safety warning is demonstrated.
TRANSPORTATION RESEARCH RECORD
(2022)
Article
Geochemistry & Geophysics
Keyan Chen, Wenyuan Li, Jianqi Chen, Zhengxia Zou, Zhenwei Shi
Summary: Remote sensing scene classification is a challenging task, but recent advances in convolutional neural networks (CNNs) have improved accuracy. However, handling resolution variations in remote sensing images is still challenging for CNNs. In this letter, we propose a novel scene classification method that leverages implicit neural representations (INRs) to adapt to scale and resolution changes. By converting the images into continuous functions and performing classification within the function space, our method achieves comparable accuracy to deep CNNs while exhibiting better adaptability to resolution changes.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2023)
Article
Geochemistry & Geophysics
Liqin Liu, Zhengxia Zou, Zhenwei Shi
Summary: This article introduces a novel method for modeling nonlinear spectral mixtures based on implicit neural representations (INRs), and proposes a new method for hyperspectral image (HSI) synthesis. The proposed method adaptively implements different mixture models for each pixel according to their spectral signature and surrounding environment, producing accurate and physically meaningful HSIs. It also generates subpixel-level spectral abundance and the solar atmosphere signature as by-products.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Engineering, Environmental
Shaofu Huang, Keyan Chen, Xiangyu Chen, Hanpeng Liao, Raymond Jianxiong Zeng, Shungui Zhou, Man Chen
Summary: This work demonstrated a biophotoelectrochemical pathway that enhances soil denitrification by directly exciting soil. The investigation of sunlight's influence on soil denitrification has mainly focused on plant photosynthesis-mediated processes. However, this study discovered a new pathway, the biophotoelectrochemical process, which greatly accelerates soil denitrification by generating more electron sources through photoinduced ferrous substrates and photoelectrons.
ENVIRONMENTAL SCIENCE & TECHNOLOGY
(2023)
Article
Geochemistry & Geophysics
Keyan Chen, Wenyuan Li, Sen Lei, Jianqi Chen, Xiaolong Jiang, Zhengxia Zou, Zhenwei Shi
Summary: A highly applicable super-resolution (SR) framework called FunSR is proposed, which handles different magnifications with a unified model. FunSR shows state-of-the-art performance on both fixed- and continuous-magnification settings and provides many friendly applications due to its unified nature.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Geochemistry & Geophysics
Hao Chen, Haotian Zhang, Keyan Chen, Chenyao Zhou, Song Chen, Zhengxia Zou, Zhenwei Shi
Summary: In this paper, a method for change detection in remote sensing images with different resolutions is proposed. The model is trained using synthesized samples to consistently predict high-resolution results for varying resolution differences. Experimental results demonstrate that the proposed method outperforms other methods on multiple datasets.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Geochemistry & Geophysics
Chenyang Liu, Rui Zhao, Jianqi Chen, Zipeng Qi, Zhengxia Zou, Zhenwei Shi
Summary: Remote sensing image change captioning (RSICC) is a novel task that aims to describe the differences between bitemporal images by natural language. This article decouples the task into two issues and utilizes prompt learning to generate captions with pretrained large language models (LLMs). Extensive experiments show that the method is effective and achieves state-of-the-art performance.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Geochemistry & Geophysics
Jianqi Chen, Keyan Chen, Hao Chen, Wenyuan Li, Zhengxia Zou, Zhenwei Shi
Summary: The article introduces an asynchronous contrastive learning-based method for effective fine-grained ship classification in remote sensing images. The method, called Push-and-Pull Network (P(2)Net), separates images using a dual-branch network and aggregates them into subclasses using an integration module. Experimental results demonstrate the effectiveness of the proposed method.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Geochemistry & Geophysics
Liqin Liu, Wenyuan Li, Zhenwei Shi, Zhengxia Zou
Summary: This article proposes a deep learning-based method for generating high-resolution hyperspectral images and subpixel ground-truth annotations from RGB images. The method takes into consideration the imaging mechanism and spectral mixing, and employs a deep generative network along with a spectral library to generate synthetic spectral data. The addition of spatial and spectral discriminative networks improves the accuracy of the synthetic output.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)