Article
Agriculture, Multidisciplinary
Anna Fabijanska, Malgorzata Danek, Joanna Barniak
Summary: This paper introduces a convolutional neural network method for automatic tree species identification from scanned wood core images, achieving high accuracy in wood patch classification and wood core classification tasks. The model outperformed the state-of-the-art methods and the study also analyzed the impact of model parameters and training settings on performance to ensure the highest recognition rates.
COMPUTERS AND ELECTRONICS IN AGRICULTURE
(2021)
Article
Chemistry, Multidisciplinary
Meng Zhu, Jincong Wang, Achuan Wang, Honge Ren, Mahmoud Emam
Summary: This study successfully increased the accuracy of identifying wood species by establishing a wood microscopic image dataset and introducing deep learning models, providing convenient conditions for further research on the microscopic characteristics of wood cells.
APPLIED SCIENCES-BASEL
(2021)
Review
Forestry
Jose Luis Silva, Rui Bordalo, Jose Pissarra, Paloma de Palacios
Summary: Wood identification is an important tool in various fields. Computer vision-based technology provides a fast and accurate method for wood identification, but its application in fields like cultural heritage is still limited.
Article
Oncology
Menghan Liu, Shuai Zhang, Yanan Du, Xiaodong Zhang, Dawei Wang, Wanqing Ren, Jingxiang Sun, Shiwei Yang, Guang Zhang
Summary: The combination of deep learning and multi-modal imaging plays a significant role in diagnosing breast cancer subtypes and assisting doctors in selecting personalized treatment plans.
FRONTIERS IN ONCOLOGY
(2023)
Article
Environmental Sciences
Jun Qin, Biao Wang, Yanlan Wu, Qi Lu, Haochen Zhu
Summary: This paper introduces a new network, SCANet, based on UAV multi-spectral remote sensing images to identify pine nematode disease. The proposed method achieved an overall accuracy rate of 79% with high precision and recall values, outperforming other existing methods. It provides a fast, precise, and practical approach for identifying nematode disease and supports the surveillance and control of this destructive disease effectively.
Article
Engineering, Multidisciplinary
Gonul Sakalli, Hasan Koyuncu
Summary: This study utilizes thermal image analysis to distinguish different conditions of asynchronous motors and transformers. By evaluating deep learning algorithms on 20 different situations, it is found that all tested architectures achieve 100% accuracy.
Article
Agronomy
Zahid Ullah, Najah Alsubaie, Mona Jamjoom, Samah H. Alajmani, Farrukh Saleem
Summary: With the increasing demand for tomatoes worldwide, there is a need to detect and prevent tomato diseases. This study proposes a hybrid deep learning model for accurately detecting tomato leaf diseases through leaf images. The proposed model achieved a success rate of 99.92% in accurately detecting tomato leaf diseases, demonstrating its reliability as an automatic detector for tomato plant diseases.
Review
Biochemical Research Methods
Sung-Wook Hwang, Junji Sugiyama
Summary: The advancements in computer vision and machine learning have revolutionized scientific disciplines and created a new research field in wood science known as computer vision-based wood identification. Research has reviewed mainstream studies using machine learning procedures to familiarize wood scientists with this area and help them choose appropriate techniques in wood science.
Article
Engineering, Geological
Diyuan Li, Zida Liu, Quanqi Zhu, Chenxi Zhang, Peng Xiao, Jinyin Ma
Summary: This paper presents a quantitative method based on SEM images and deep learning to identify the mesoscopic failure mechanism of granite, which can determine the distribution of tensile and shear fractures on failure surfaces. The developed AlexNet models achieved high accuracy (96-98%) in identifying tensile and shear fracture surfaces. The results verified the application of the proposed method and its usefulness in understanding the failure mechanism of rock under uniaxial compression.
ROCK MECHANICS AND ROCK ENGINEERING
(2023)
Article
Agriculture, Dairy & Animal Science
Taejun Lee, Youngjun Na, Beob Gyun Kim, Sangrak Lee, Yongjun Choi
Summary: This study successfully identified Hanwoo cattle using a deep-learning model and muzzle images. By taking images of the same individuals at different times, overfitting models were avoided. Among multiple transfer-learned models, the small version of Efficientnet v2 with Lion optimizer achieved the best accuracy. Muzzle patterns were demonstrated to have potential as a key for individual cattle identification.
Article
Computer Science, Artificial Intelligence
Gustavo Z. Felipe, Jacqueline N. Zanoni, Camila C. Sehaber-Sierakowski, Gleison D. P. Bossolani, Sara R. G. Souza, Franklin C. Flores, Luiz E. S. Oliveira, Rodolfo M. Pereira, Yandre M. G. Costa
Summary: The research utilized pattern recognition and machine learning techniques to evaluate the impact of chronic degenerative diseases on EGC, demonstrating an effective method to distinguish healthy cells from diseased cells.
NEURAL COMPUTING & APPLICATIONS
(2021)
Article
Environmental Sciences
Qi Lv, Qian Li, Kai Chen, Yao Lu, Liwen Wang
Summary: This paper proposes a ground-based cloud image classification method based on contrastive self-supervised learning, which improves the accuracy of cloud type classification by reducing the dependence on labeled samples. Experimental results confirm the effectiveness of the method, providing inspiration and technical reference for the analysis and processing of meteorological remote sensing data.
Article
Environmental Sciences
Binbin Wang, Guijun Yang, Hao Yang, Jinan Gu, Sizhe Xu, Dan Zhao, Bo Xu
Summary: This study utilizes UAV remote sensing technology and a deep learning algorithm to improve the detection and counting of maize tassels, addressing the challenges posed by the complex field environment. The improved RetinaNet model demonstrates significant advancements in accuracy compared to other mainstream target detection models.
Article
Biochemistry & Molecular Biology
Zihan Yang, Hongming Pan, Jianwei Shang, Jun Zhang, Yanmei Liang
Summary: Early detection and diagnosis of oral cancer are crucial for a better prognosis, but accurate and automatic identification using current technologies is challenging. This study aims to assess deep-learning-based algorithms for optical coherence tomography (OCT) images in assisting clinicians with oral cancer screening and diagnosis. The performance of three convolutional neural networks (CNNs) was evaluated using four metrics, and compared with machine learning approaches. The results demonstrate that CNNs outperform machine-learning-based methods, with a classification accuracy of up to 96.76%. Additionally, the visualization of lesions in OCT images and the interpretability of the model for distinguishing different oral tissues were evaluated. This study proves the potential of deep-learning-based automatic identification algorithm for OCT images in providing decision support for effective oral cancer screening and diagnosis.
Article
Computer Science, Information Systems
Mao Xiao, Chun Guo, Guowei Shen, Yunhe Cui, Chaohui Jiang
Summary: This paper presents a malware classification method based on PE files, using a new visualization method and deep learning technology to improve the accuracy and efficiency of malware classification.
COMPUTERS & SECURITY
(2021)
Article
Computer Science, Software Engineering
Suren Deepak Rajasekaran, Hao Kang, Martin Cadik, Eric Galin, Eric Guerin, Adrien Peytavie, Pavel Slavik, Bedrich Benes
Summary: This article introduces a method for perceptually evaluating the realism of terrain models, by categorizing real terrains and generating synthetic ones, as well as analyzing the impact and importance of features on perceived realism. Through quantitative evaluation and neural network transfer experiments, the influence of terrain features is validated, and a new metric for assessing terrain realism (PTRM) is proposed.
ACM TRANSACTIONS ON APPLIED PERCEPTION
(2022)
Article
Computer Science, Software Engineering
Till Niese, Soren Pirk, Matthias Albrecht, Bedrich Benes, Oliver Deussen
Summary: This paper introduces a procedural model for vegetation placement in urban landscapes. The model takes into account the city's geometry and determines plausible plant positions based on the structural and functional zones. The model can be directly used or learned from satellite images and land register data. The effectiveness of the framework is demonstrated through examples and validated through user studies and design sessions with expert users.
ACM TRANSACTIONS ON GRAPHICS
(2022)
Article
Computer Science, Software Engineering
Mathieu Gaillard, Vojtech Krs, Giorgio Gori, Radomir Mech, Bedrich Benes
Summary: ADPM introduces automatic differentiable procedural modeling, providing users with an interactive way to model 3D objects while preserving the procedural representation. This method gives precise control over the resulting model, comparable to non-procedural interactive modeling.
COMPUTER GRAPHICS FORUM
(2022)
Article
Engineering, Civil
Yang Liu, Fanyou Wu, Cheng Lyu, Xin Liu, Zhiyuan Liu
Summary: This paper investigates the effective embedding of users' personalized travel behaviors to vectors, proposing an improved method named Behavior2vector tailored for this purpose. Through machine learning model design and analysis of various factors affecting travel mode choices, the impact of user behavior representation on intelligent transportation systems is explored and validated using travel big data. The study also compares existing graph embedding methods and discusses their advantages and disadvantages.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2022)
Article
Computer Science, Software Engineering
Adnan Firoze, Bedrich Benes, Daniel Aliaga
Summary: We propose a vision-based algorithm that combines satellite imagery, pattern recognition, procedural modeling, and deep learning to accurately locate trees in urban areas. By utilizing satellite snapshots and vegetation clustering, as well as employing GAN networks for procedural tree planting, our algorithm achieves high tree count accuracies in four different cities.
Article
Construction & Building Technology
Ju An Park, Xiaoyu Liu, Chul Min Yeum, Shirley J. Dyke, Max Midwinter, Jongseong Choi, Zhiwei Chu, Thomas Hacker, Bedrich Benes
Summary: This paper introduces a comprehensive classification schema and a multi-output DCNN model for rapid postearthquake image classification. The performance of the proposed multi-output model was validated and shown to outperform other models. The model has been deployed to a web-based platform for organizing earthquake reconnaissance images.
JOURNAL OF PERFORMANCE OF CONSTRUCTED FACILITIES
(2022)
Article
Computer Science, Software Engineering
Tomas Polasek, Martin Cadik, Yosi Keller, Bedrich Benes
Summary: We present Vision UFormer (ViUT), a novel deep neural network for monocular depth estimation over long range. ViUT takes an RGB image as input and generates a depth image where each pixel represents the absolute distance of the corresponding object in the scene. It incorporates a Transformer encoder, a ResNet decoder, and UNet style skip connections, and is trained on 1M images from ten different datasets. The results show that ViUT performs well on normalized relative distances and short-range classical datasets, and successfully estimates absolute long-range depth in meters.
COMPUTERS & GRAPHICS-UK
(2023)
Article
Agriculture, Multidisciplinary
Mathieu Gaillard, Bedrich Benes, Michael C. Tross, James C. Schnable
Summary: We propose a new method for solving the joint problem of correspondence and triangulation of points from multiple calibrated perspective views. This method has been successfully applied to counting the number of leaves on plants photographed from multiple angles. Our algorithm is robust to noise and occlusion, and can infer occluded points by reasoning on the 3D geometry of the scene. It can handle a large number of points reconstructed from a reduced set of views.
COMPUTERS AND ELECTRONICS IN AGRICULTURE
(2023)
Article
Computer Science, Software Engineering
Guillaume Cordonnier, Guillaume Jouvet, Adrien Peytavie, Jean Braun, Marie-Paule Cani, Bedrich Benes, Eric Galin, Eric Guerin, James Gain
Summary: We present a novel solution for simulating the formation and evolution of glaciers, as well as their erosive effects, in the context of glacial and inter-glacial cycles. Our solution includes a fast and accurate deep learning-based estimation method for high-order ice flows, as well as a new multi-scale advection scheme to handle the different time scales of glacier equilibrium and terrain erosion. By combining these methods, we are able to accurately model the formation of various terrain features, including U-shaped and hanging valleys, fjords, and glacial lakes, which were not adequately represented in previous computer graphics models.
ACM TRANSACTIONS ON GRAPHICS
(2023)
Article
Computer Science, Software Engineering
Bosheng Li, Jonathan Klein, Dominik L. Michels, Bedrich Benes, Soren Pirk, Wojtek Palubicki
Summary: Computer graphics has focused on generating realistic models of trees and plants. Existing methods use procedural modeling algorithms to create branching structures for trees, but often neglect to model the root system. In this paper, we introduce a physically-plausible soil model, a novel developmental procedural model for tree roots, and long-distance signaling to coordinate tree development, enabling the generation of trees with their root systems for the first time.
ACM TRANSACTIONS ON GRAPHICS
(2023)
Article
Computer Science, Software Engineering
Simon Perche, Adrien Peytavie, Bedrich Benes, Eric Galin, Eric Guerin
Summary: This article introduces a new generative network method that combines automatic terrain synthesis with authoring, providing a versatile set of authoring tools. The method can generate terrains from input sketches or existing elevation maps, and further enhance them using interactive brushes and style manipulation tools. The strength of the approach lies in the versatility and interoperability of the tools, which have been verified quantitatively and qualitatively.
COMPUTER GRAPHICS FORUM
(2023)
Article
Multidisciplinary Sciences
Yang Liu, Fanyou Wu, Zhiyuan Liu, Kai Wang, Feiyue Wang, Xiaobo Qu
Summary: Language models can be used to learn representations of entities beyond language, such as human behaviors. This study proposes a novel approach based on language models to optimize delivery routes by learning from drivers' historical experiences. Experimental results on real-world data demonstrate the effectiveness and scalability of the proposed approach.
Proceedings Paper
Computer Science, Artificial Intelligence
Yichen Sheng, Yifan Liu, Jianming Zhang, Wei Yin, A. Cengiz Oztireli, He Zhang, Zhe Lin, Eli Shechtman, Bedrich Benes
Summary: This article introduces a novel geometry representation called Pixel Height, which can be calculated in various ways and has the advantages of significantly improving the quality and controllability of shadow generation.
COMPUTER VISION, ECCV 2022, PT XXIII
(2022)
Proceedings Paper
Engineering, Aerospace
Fan Yang, Jiansong Zhang, Bedrich Benes
Summary: This paper provides a survey of the state-of-the-art works on building fire simulations from 2015 to 2020, revealing a focus on high-rise and public buildings in the research. It also summarizes the latest technological advances in this field.
CONSTRUCTION RESEARCH CONGRESS 2022: COMPUTER APPLICATIONS, AUTOMATION, AND DATA ANALYTICS
(2022)
Review
Computer Science, Interdisciplinary Applications
Jose Daniel Azofeifa, Julieta Noguez, Sergio Ruiz, Jose Martin Molina-Espinosa, Alejandra J. Magana, Bedrich Benes
Summary: This document presents a systematic review of Multimodal Human-Computer Interaction and analyzes 112 relevant research works out of the initial 406 articles. It identifies virtual reality (VR) and haptics as the most widely used technologies in various domains, suggesting the potential for future applications that combine VR and haptic interaction.
