Article
Computer Science, Information Systems
Bujar Fetai, Dejan Grigillo, Anka Lisec
Summary: The goal of this research was to develop an approach to detect visible land boundaries and revise existing cadastral data using deep learning. The results showed good detection quality but lower accuracy compared to manual methods.
ISPRS INTERNATIONAL JOURNAL OF GEO-INFORMATION
(2022)
Article
Engineering, Electrical & Electronic
Roohollah Amiri, Srinivas Yerramalli, Taesang Yoo, Mohammed Hirzallah, Marwen Zorgui, Rajat Prakash, Xiaoxia Zhang
Summary: This paper proposes a novel sensing solution for representing an RF-environment and addresses practical challenges and wireless propagation phenomena. It utilizes offline data collection and an iterative process to locate virtual anchors and trains machine learning models to predict the dominant multipath components of the wireless channel. These models are used to improve positioning accuracy in challenging indoor environments through multipath assisted positioning.
IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS
(2023)
Article
Environmental Studies
Chryssy Potsiou, Nikolaos Doulamis, Nikolaos Bakalos, Maria Gkeli, Charalabos Ioannidis, Selena Markouizou
Summary: This paper presents an innovative research aiming to propose a low-cost and reliable method to support the registration of informal, multi-story and unregistered constructions in self-made cities. The method includes an Indoor Positioning System combined with a machine learning algorithm and a 3D cadastral mapping mobile application.
Article
Environmental Sciences
Getachew Workineh Gella, Lorenz Wendt, Stefan Lang, Dirk Tiede, Barbara Hofer, Yunya Gao, Andreas Braun
Summary: This study investigates the use of a deep convolutional neural network-based Mask R-CNN model for dwelling extractions in IDP/refugee settlements. The model was trained using transfer learning from historical images, and showed better performance compared to training from scratch.
Article
Remote Sensing
Maxwell Jong, Kaiyu Guan, Sibo Wang, Yizhi Huang, Bin Peng
Summary: Field boundary data is essential for digital agricultural services and research, and adversarial training can significantly improve the quality and performance of field boundary prediction.
INTERNATIONAL JOURNAL OF APPLIED EARTH OBSERVATION AND GEOINFORMATION
(2022)
Article
Health Care Sciences & Services
Bo-Kyeong Kang, Yelin Han, Jaehoon Oh, Jongwoo Lim, Jongbin Ryu, Myeong Seong Yoon, Juncheol Lee, Soorack Ryu
Summary: This study developed and validated an automatic segmentation algorithm for the delineation of ten wrist bones using a convolutional neural network. The Fine Mask R-CNN model showed improved performance in accurately segmenting wrist bones compared to traditional methods. The highest performance was observed in the distal radius, while the lowest was in the trapezoid, highlighting the efficacy of the proposed model for automatic segmentation.
JOURNAL OF PERSONALIZED MEDICINE
(2022)
Article
Oncology
Julius C. Holzschuh, Michael Mix, Juri Ruf, Tobias Hoelscher, Joerg Kotzerke, Alexis Vrachimis, Paul Doolan, Harun Ilhan, Ioana M. Marinescu, Simon K. B. Spohn, Tobias Fechter, Dejan Kuhn, Peter Bronsert, Christian Gratzke, Radu Grosu, Sophia C. Kamran, Pedram Heidari, Thomas S. C. Ng, Arda Koenik, Anca-Ligia Grosu, Constantinos Zamboglou
Summary: A deep learning model was created to accurately delineate the intraprostatic gross tumor volume (GTV) in PSMA-PET, achieving fast segmentation with high diagnostic accuracy comparable to manual experts.
RADIOTHERAPY AND ONCOLOGY
(2023)
Article
Geosciences, Multidisciplinary
Phuong Thao Thi Ngo, Mahdi Panahi, Khabat Khosravi, Omid Ghorbanzadeh, Narges Kariminejad, Artemi Cerda, Saro Lee
Summary: In this study, a national-scale landslide susceptibility mapping of Iran was conducted using recurrent neural network (RNN) and convolutional neural network (CNN) algorithms. RNN algorithm outperformed CNN algorithm in both training and testing phases, with 6% and 14% of Iran's land area being highly susceptible to future landslide events. Additionally, 31% of cities in Iran are located in high and very high landslide susceptibility areas.
GEOSCIENCE FRONTIERS
(2021)
Article
Computer Science, Artificial Intelligence
Abdolrahman Peimankar, Sadasivan Puthusserypady
Summary: This study proposes a deep learning model for heartbeat segmentation, which combines convolutional neural network and long short-term memory model to analyze ECG signals in real-time, achieving high sensitivity and precision.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Environmental Sciences
Wei Huang, Zeping Liu, Hong Tang, Jiayi Ge
Summary: This paper introduces a novel method for sequentially delineating exterior and interior contours of rooftops with holes from VHR aerial images, integrating semantic segmentation and polygon delineation. By using a convolutional recurrent neural network, this method effectively delineates rooftops with both one and multiple polygons, outperforming existing methods in terms of visual results and six statistical indicators.
Article
Geochemistry & Geophysics
Shiqing Wei, Tao Zhang, Shunping Ji
Summary: In this article, a concentric loop convolutional neural network (CLP-CNN) method is proposed for the automatic segmentation of building boundaries from remote-sensing images. The experiments show that the proposed method performs well on building datasets and generic object boundary delineation tests.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Engineering, Environmental
Hossein Hamedi, Ali Asghar Alesheikh, Mahdi Panahi, Saro Lee
Summary: Using deep learning algorithms including CNN and LSTM, landslide prone areas were identified in Ardabil province, Iran. The LSTM model showed slightly better performance compared to the CNN model, but both models have close performance with acceptable accuracy. AUC values for CNN and LSTM models were 0.821 and 0.832, respectively, indicating the effectiveness of the models in landslide susceptibility mapping.
STOCHASTIC ENVIRONMENTAL RESEARCH AND RISK ASSESSMENT
(2022)
Article
Engineering, Environmental
Ahmed M. Youssef, Biswajeet Pradhan, Abhirup Dikshit, Mohamed M. Al-Katheri, Saleh S. Matar, Ali M. Mahdi
Summary: This study compares the accuracy of support vector machine, one-dimensional convolutional neural network, and two-dimensional convolutional neural network models in landslide susceptibility mapping in the Asir Region, Saudi Arabia. The experimental results show that the CNN-1D and CNN-2D models are more accurate than the conventional machine learning model (SVM) in predicting landslide spatial distribution, with CNN-2D model performing the best.
BULLETIN OF ENGINEERING GEOLOGY AND THE ENVIRONMENT
(2022)
Article
Chemistry, Multidisciplinary
Bo Wang, Fan Shi, Haiyang Zheng
Summary: With the proliferation of internet technology, the rise of illicit websites such as gambling and pornography has become a serious concern due to the threats they pose to people's well-being and financial security. Current governance measures rely on manual detection, but the need for effective and efficient solutions is urgent. This paper proposes a method that utilizes web mapping engine big data to perform unsupervised multimodal clustering for the discovery of illicit websites, achieving high accuracy in identification and classification.
APPLIED SCIENCES-BASEL
(2023)
Article
Geochemistry & Geophysics
Adnan Farooq, Xiuping Jia, Jiankun Hu, Jun Zhou
Summary: Automatic weed monitoring and classification are crucial for site-specific weed management. A deep learning-based convolutional neural network can effectively discriminate weed species by learning spectral, spatial, and structural features. Transfer learning with a partial CNN model shows promising results for handling target datasets with limited training samples and varying spatial resolutions.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Editorial Material
Geography, Physical
Michael Ying Yang, Loic Landrieu, Devis Tuia, Charles Toth
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING
(2021)
Article
Geography, Physical
Ye Lyu, Michael Ying Yang, George Vosselman, Gui-Song Xia
Summary: Video object detection is less researched compared to object detection in images due to shortage of labelled video datasets. Frames in a video clip are highly correlated, requiring more video labels for good data variation. Propose to improve the performance of an image object detector by augmenting it with a class-agnostic convolutional regression tracker.
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING
(2021)
Article
Computer Science, Artificial Intelligence
Simone Borsci, Ville V. Lehtola, Francesco Nex, Michael Ying Yang, Ellen-Wien Augustijn, Leila Bagheriye, Christoph Brune, Ourania Kounadi, Jamy Li, Joao Moreira, Joanne Van der Nagel, Bernard Veldkamp, Duc Le, Mingshu Wang, Fons Wijnhoven, Jelmer M. Wolterink, Raul Zurita-Milla
Summary: The article reviews the EU Commission's whitepaper on Artificial Intelligence and highlights potential conflicts with current societal, technical, and methodological constraints. The lack of a coherent EU vision and methods to support sustainable AI diffusion are identified as main obstacles. The article recommends complementary rules and compensatory mechanisms to avoid market fragmentation, as well as research to address technical and methodological open questions for the sustainable development of human-AI co-action.
Article
Geography, Physical
Yaping Lin, George Vosselman, Michael Ying Yang
Summary: This paper proposes a weakly supervised approach for semantic segmentation of ALS point clouds, achieving efficient and accurate segmentation on large-scale datasets through the use of pseudo labels and improved network structures. Experimental results demonstrate that the method outperforms traditional approaches in terms of performance.
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING
(2022)
Article
Engineering, Mechanical
Guizhong Fu, Shukai Jia, Wenbin Zhu, Jiangxin Yang, Yanlong Cao, Michael Ying Yang, Yanpeng Cao
Summary: This paper proposes a multi-light source illumination/acquisition system and a multi-stream CNN model for high-accuracy surface defect classification on highly reflective metal. By fusing features extracted from multi-light source illuminated images, more accurate recognition results can be generated. In addition, the authors also propose individual stream deep supervision and channel attention-based feature re-calibration techniques.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2022)
Article
Environmental Sciences
Ning Zhang, Francesco Nex, George Vosselman, Norman Kerle
Summary: This research focuses on using deep learning to detect victims in disaster debris, proposes a method to generate harmonious composite images for training, and significantly improves detection accuracy.
Article
Engineering, Civil
Fashuai Li, Zhize Zhou, Jianhua Xiao, Ruizhi Chen, Matti Lehtomaki, Sander Oude Elberink, George Vosselman, Juha Hyyppa, Yuwei Chen, Antero Kukko
Summary: This paper presents an improved framework for instance-aware semantic segmentation of road furniture in mobile laser scanning data. The framework detects road furniture, decomposes them into poles and components, extracts instance information, and classifies the components using a classifier and DenseCRF. The combination of random forest with DenseCRF achieves high overall accuracies.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Jian Ding, Nan Xue, Gui-Song Xia, Xiang Bai, Wen Yang, Michael Ying Yang, Serge Belongie, Jiebo Luo, Mihai Datcu, Marcello Pelillo, Liangpei Zhang
Summary: This paper presents a large-scale DOTA dataset for object detection in aerial images, along with comprehensive baselines and a code library. The dataset and evaluations provided can facilitate the design of robust algorithms and reproducible research in the field of object detection in aerial images.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2022)
Article
Computer Science, Artificial Intelligence
Yanpeng Cao, Xing Luo, Jiangxin Yang, Yanlong Cao, Michael Ying Yang
Summary: This paper proposes a novel multispectral pedestrian detection method that generates highly discriminative features by aggregating human-related clues in multispectral images. By performing cross-modal feature aggregation and pixel-level detection fusion, the proposed method achieves improved accuracy in pedestrian detection.
INFORMATION FUSION
(2022)
Article
Geography, Physical
Yunshuang Yuan, Hao Cheng, Michael Ying Yang, Monika Sester
Summary: This study introduces a probabilistic model called GevBEV for accurately capturing the uncertainties of the perception system in autonomous driving. Experimental results demonstrate that the model provides reliable uncertainty quantification and performs well in cooperative perception scenarios.
ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Wentong Liao, Kai Hu, Michael Ying Yang, Bodo Rosenhahn
Summary: Text-to-image synthesis aims to generate photo-realistic images that are semantically consistent with the text descriptions. Existing methods have limitations in capturing fine-grained details. In this paper, a novel framework is proposed to improve the visual fidelity and alignment with input text description by introducing a semantic-spatial aware block.
2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2022)
(2022)
Proceedings Paper
Geography, Physical
A. Maiti, S. J. Oude Elberink, G. Vosselman
Summary: This paper investigates the impact of label noise on the performance of deep learning models in semantic segmentation. Experimental results show that label noise decreases the accuracy of the model, and different classes respond differently to label noise.
XXIV ISPRS CONGRESS IMAGING TODAY, FORESEEING TOMORROW, COMMISSION II
(2022)
Proceedings Paper
Geography, Physical
N. Zhang, F. Nex, G. Vosselman, N. Kerle
Summary: This paper addresses the issue of deep detection networks in detecting buried victims. By generating realistic images and using an unsupervised generative adversarial network for harmonization, the accuracy of victim detection can be effectively improved.
XXIV ISPRS CONGRESS: IMAGING TODAY, FORESEEING TOMORROW, COMMISSION III
(2022)
Article
Engineering, Electrical & Electronic
Yang Long, Gui-Song Xia, Shengyang Li, Wen Yang, Michael Ying Yang, Xiao Xiang Zhu, Liangpei Zhang, Deren Li
Summary: This article discusses how to efficiently prepare a suitable benchmark dataset for remote sensing (RS) image interpretation, presenting general guidances on creating benchmark datasets and providing an example of a Million Aerial Image Dataset. It also addresses challenges and perspectives in RS image annotation to facilitate research in benchmark dataset construction.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
(2021)