Article
Computer Science, Artificial Intelligence
Y. S. Gan, Weihao Chen, Wei-Chuen Yau, Ziyun Zou, Sze-Teng Liong, Shih-Yuan Wang
Summary: Image-based 3D reconstruction from a single-view image is critical and fundamental, but faces challenges like self-occlusion and lack of object information. This paper proposes a new and simple, yet powerful framework that improves the quality of the generated point cloud from a single-view image. The proposed algorithm demonstrates its robustness and effectiveness compared to state-of-the-art 3D reconstruction methods.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Meihua Zhao, Gang Xiong, MengChu Zhou, Zhen Shen, Fei-Yue Wang
Summary: The paper proposes a method called 3D-RVP, which can reconstruct complete and accurate 3D geometry from a single depth view, addressing the issue of difficulty in improving the resolution of 3D object reconstruction technology. By combining a 3D encoder-decoder network and a point prediction network, the method can output high-resolution 3D models, outperforming existing technologies.
Article
Computer Science, Information Systems
Lei Zhu, Shanmin Wang, Zengqun Zhao, Xiang Xu, Qingshan Liu
Summary: This paper proposes a contextual encoder-decoder network named CED-Net for regressing UV position map in 3D face reconstruction. The network incorporates contextual information at both shape and feature levels, capturing the relationship between facial features from a spatial perspective. Experimental results show that the proposed method achieves superior results in both benchmarks.
MULTIMEDIA SYSTEMS
(2022)
Article
Environmental Sciences
Yang Li, Hui Lu, Qi Liu, Yonghong Zhang, Xiaodong Liu
Summary: In this paper, we propose a new single-side dual-branch network (SSDBN) based on an encoder-decoder structure, which can accurately perform semantic segmentation and improve the capturing ability of semantic details.
Article
Biology
Zhongxi Qiu, Yan Hu, Jiayi Zhang, Xiaoshan Chen, Jiang Liu
Summary: This paper proposes a new attention module (FGAM) for medical image segmentation, which is simple, pluggable, and effective. It improves segmentation results by digging out the feature representation ability in the encoder and decoder features.
COMPUTERS IN BIOLOGY AND MEDICINE
(2022)
Article
Computer Science, Artificial Intelligence
Mohammad Golbabaee, Guido Buonincontri, Carolin M. Pirkl, Marion Menzel, Bjoern H. Menze, Mike Davies, Pedro A. Gomez
Summary: A novel pipeline for multi-parametric quantitative MRI image computing is proposed, utilizing compressed sensing reconstruction and deep learned quantitative inference. The approach effectively recovers accurate and consistent quantitative information through flexible generation of rich training samples in the trained model.
MEDICAL IMAGE ANALYSIS
(2021)
Article
Computer Science, Information Systems
Xiang Tian, Xuechao Liu, Tao Zhang, Jian'an Ye, Weirui Zhang, Liangliang Zhang, Xuetao Shi, Feng Fu, Zhongyu Li, Canhua Xu
Summary: Electrical impedance tomography (EIT) is a noninvasive and radiation-free imaging method. This study introduces an enhanced encoder-decoder (EED) method with an atrous spatial pyramid pooling (ASPP) module to address the issue of weak central target reconstruction in the presence of strong edge targets.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2023)
Article
Environmental Sciences
Rui Zhao, Shihong Du
Summary: In this study, a deep learning model of an encoder-decoder with a residual network (EDRN) is proposed for fusing hyperspectral and panchromatic remote sensing images. The experimental results demonstrate the superior performance of the proposed method on six real-world datasets.
Article
Computer Science, Information Systems
Aiping Yang, Chaochen Wang, Jinbin Wang, Qian Wang, Tengfei Zhang
Summary: This paper presents a novel underwater image enhancement approach based on training an end-to-end underwater image enhancement network without using any reference image. A novel encoder-decoder network structure and a set of non-reference loss functions are designed to measure the enhancement quality. The subjective and objective evaluations show that the proposed algorithm outperforms the state-of-the-art approaches.
MULTIMEDIA TOOLS AND APPLICATIONS
(2023)
Article
Computer Science, Information Systems
Aniruddha Mazumdar, Prabin Kumar Bora
Summary: This paper presents a two-stream encoder-decoder network that utilizes both high-level and low-level image features to precisely localize forged regions in manipulated images. By learning different levels of features in two streams, it improves the accuracy of detecting forged regions.
JOURNAL OF VISUAL COMMUNICATION AND IMAGE REPRESENTATION
(2022)
Article
Chemistry, Multidisciplinary
Shumpei Nemoto, Tadahaya Mizuno, Hiroyuki Kusuhara
Summary: Descriptor generation methods using latent representations of encoder-decoder models with SMILES as input have the advantage of continuity and restorability to the structure. This study investigates the learning progress of ED models to understand how they recognize the structure. The results show that compound substructures are learned early in the learning process, but structure restoration is time-consuming and can result in overestimation.
JOURNAL OF CHEMINFORMATICS
(2023)
Article
Computer Science, Artificial Intelligence
Di Wang, Yue Pan, Oguz C. Durumeric, Joseph M. Reinhardt, Eric A. Hoffman, Joyce D. Schroeder, Gary E. Christensen
Summary: This paper presents the PLOSL pulmonary image registration method, which combines the advantages of population learning and one-shot learning to achieve fast and efficient registration of lung images. By using tissue volume preserving and vesselness constraints for image matching, PLOSL is able to accurately extract lung shape features and achieve good registration results on multiple datasets.
MEDICAL IMAGE ANALYSIS
(2022)
Article
Computer Science, Hardware & Architecture
Vidhi Chaudhary, Preetpal Kaur Buttar, Manoj Kumar Sachan
Summary: Road network is crucial for urban development and various applications. This paper investigates the potential and performance of the U-Net convolution neural network architecture for road detection and achieves improved segmentation results through proposed techniques and hyperparameters.
JOURNAL OF SUPERCOMPUTING
(2022)
Article
Computer Science, Interdisciplinary Applications
Hyunoh Lee, Jinwon Lee, Hyungki Kim, Duhwan Mun
Summary: This study proposed a method to effectively reconstruct 3D CAD models using a 3D encoder-decoder network, which was trained on large-scale 3D CAD model datasets and tested on numerous parts to demonstrate high reconstruction performance with an error rate of approximately 1%.
JOURNAL OF COMPUTATIONAL DESIGN AND ENGINEERING
(2022)
Article
Computer Science, Artificial Intelligence
Junchao Zhang, Jianbo Shao, Jianlai Chen, Degui Yang, Buge Liang
Summary: The proposed deep network method for polarization image fusion with self-learned strategy and unsupervised training outperforms state-of-the-art methods in visual quality and quantitative measurement. It can be applied in military and civilian fields for camouflage, hidden targets detection, medical diagnosis, and environmental monitoring.
PATTERN RECOGNITION
(2021)
Article
Geochemistry & Geophysics
Nan Su, Zhibo Huang, Yiming Yan, Chunhui Zhao, Shuyuan Zhou
Summary: Ship detection is a major problem in satellite image analysis, especially for fast detection of ships in large-area remote-sensing images. This letter proposes an arbitrary-oriented detector based on YOLO, which can quickly locate ship positions. By reducing parameters, adding deformable convolution, and integrating rotation detection capability without angle regression, the proposed method achieves excellent accuracy and speed.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Environmental Sciences
Nan Su, Jiayue He, Yiming Yan, Chunhui Zhao, Xiangwei Xing
Summary: In this paper, a novel ship detection method called SII-Net is proposed to enhance the detection performance of small ships in SAR images. The method utilizes a channel-location attention mechanism and a high-level features enhancement module to improve the accuracy of detection. Experimental results on public datasets show that the proposed method outperforms state-of-the-art detectors, especially for small-sized targets.
Article
Environmental Sciences
Fengjiao Gao, Yiming Yan, Hemin Lin, Ruiyao Shi
Summary: This paper proposes a novel 3D semantic segmentation method called PIIE-DSA-net, which combines low-level features and deep features for more reliable feature extraction. Experimental results demonstrate that this method achieves good segmentation results on both indoor and outdoor datasets.
Article
Environmental Sciences
Chunhui Zhao, Hongjiao Liu, Nan Su, Lu Wang, Yiming Yan
Summary: Object tracking using RGB images may fail when the object's color is similar to the background. Hyperspectral images provide more spectral information for RGB-based trackers, but there is currently no fusion tracking algorithm for hyperspectral and RGB images. The proposed reliability-guided aggregation network (RANet) combines hyperspectral and RGB information to improve tracking performance, with the RANet achieving the best performance accuracy among the tested trackers.
Article
Environmental Sciences
Chunhui Zhao, Hongjiao Liu, Nan Su, Congan Xu, Yiming Yan, Shou Feng
Summary: This study proposes a Transformer-based multimodality information transfer network (TMTNet) to improve hyperspectral object tracking by efficiently transferring multimodality data information composed of RGB and hyperspectral data. Two subnetworks are constructed to transfer multimodality fusion information and robust RGB visual information, respectively. The proposed TMTNet tracker outperforms advanced trackers, demonstrating its effectiveness.
Article
Environmental Sciences
Chunhui Zhao, Wenxuan Wang, Yiming Yan, Nan Su, Shou Feng, Wei Hou, Qingyu Xia
Summary: A novel object-level building-matching method using cross-dimensional data is proposed in this work. The method utilizes a plug-and-play Joint Descriptor Extraction Module (JDEM) to extract three-dimensional shape information from object-level remote sensing data for matching. The effectiveness of the method is verified using a cross-dimensional object-level data set.
Article
Environmental Sciences
Yiming Yan, Weikun Zhou, Nan Su, Chi Zhang
Summary: In this paper, a novel 3D surface reconstruction method called UniRender is proposed to address the challenges in accurately capturing the appearance and geometry of scenes in remote sensing environments using neural rendering. UniRender combines the strengths of surface and volume rendering, incorporates photometric consistency constraints, and improves the sampling strategy to achieve high-quality 3D surface reconstruction.
Article
Geochemistry & Geophysics
Jiayue He, Nan Su, Congan Xu, Yanping Liao, Yiming Yan, Chunhui Zhao, Wei Hou, Shou Feng
Summary: This article proposes a cross-modality feature transfer (CMFT) method to enhance feature representations in SAR modality by transferring knowledge from RGB modality. Through a multilevel modality alignment network (MMAN), a hard-sample supervision module (HSM), and a feature complementary module (FCM), the non-intuitive feature representations in SAR imaging are addressed. Experimental results demonstrate the superiority of CMFT method in SAR ship detection performance.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2023)
Article
Geochemistry & Geophysics
Chunhui Zhao, Minghua Wang, Nan Su, Yiming Yan, Shou Feng
Summary: Introduces a new dynamic-static transformer style network (DSTNet) that aims to reduce the difference between the SR-Recovered image and the HR image and obtain the feature representation of both images. Static and dynamic context information is extracted using CNN and Transformer, respectively, to enhance the feature representation. Different normalization strategies are designed in different depths of the network to obtain invariant information between the SR-Recovered image and the HR image. Experimental results show improvements in rank-5 and mAP performances by 3% and 3.6%, respectively, on datasets with large resolution differences.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2023)
Article
Geochemistry & Geophysics
Yiming Yan, Zilu Wang, Congan Xu, Nan Su
Summary: This paper proposes an implicit modeling framework called GEOP-Net embedded with high-dimensional geometric features for shape reconstruction of buildings based on LiDAR point clouds. Experimental results show that the proposed method has better accuracy than existing methods and provides a new research idea for building reconstruction.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2023)