Article
Geochemistry & Geophysics
Donato Amitrano, Raffaella Guida, Pasquale Iervolino
Summary: The proposed methodology introduces a new approach for unsupervised change detection in vegetation canopy, achieving superior results compared to literature methods in agriculture experiments. While maintaining comparable detection accuracy in cases of deforestation, it significantly reduces the number of false deforestation patterns. The architecture's main characteristics are robustness and lack of supervision, making it well-suited for operational scenarios.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2021)
Article
Forestry
Alba Garcia-Cimarras, Jose Antonio Manzanera, Ruben Valbuena
Summary: This study analyzed vegetation changes and canopy fuel types using LiDAR data, revealing relationships between stand structure and ecological factors, as well as trends in fuel types over a 6-year period, and assessed fire risk accordingly.
Article
Geochemistry & Geophysics
Genc Hoxha, Seloua Chouaf, Farid Melgani, Youcef Smara
Summary: This article introduces a system for describing changes in bitemporal images through change sentences to provide user-friendly interpretation. The system utilizes convolutional neural networks to extract features from bitemporal images and uses recurrent neural networks or support vector machines to generate coherent change descriptions, achieving promising experimental results.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Multidisciplinary Sciences
Jun Ruan, Zhikui Zhu, Chenchen Wu, Guanglu Ye, Jingfan Zhou, Junqiu Yue
Summary: With the rapid development of whole-slide digital scanners for histopathology, computer-assisted digital pathology image analysis has gained clinical attention, impacting the working style of pathologists. This paper presents a novel lightweight detection framework for automatic tumor detection in whole-slide histopathology images, achieving excellent performance in pixel-level detection with improved methods and models.
Article
Environmental Sciences
Kaiyu Zhang, Xikai Fu, Xiaolei Lv, Jili Yuan
Summary: This study presents a novel multitemporal building change detection framework that can extract change frequency and change moment information from time-series SAR images. By combining different methods, it effectively identifies changed building objects.
Article
Computer Science, Artificial Intelligence
Weiming Hu, Qiang Wang, Li Zhang, Luca Bertinetto, Philip H. S. Torr
Summary: In this article, SiamMask, a framework for real-time visual object tracking and video object segmentation, using the same simple method, is introduced. The offline training procedure of popular fully-convolutional Siamese approaches is improved by adding a binary segmentation task. Once the offline training is completed, SiamMask can perform visual object tracking and segmentation at high frame-rates with only a single bounding box for initialization. The framework can also handle multiple object tracking and segmentation by re-using the multi-task model in a cascaded fashion.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Robotics
Federico Ceola, Elisa Maiettini, Giulia Pasquale, Giacomo Meanti, Lorenzo Rosasco, Lorenzo Natale
Summary: In this study, we focus on designing a fast instance segmentation learning pipeline for robotic applications. The pipeline utilizes a pre-trained CNN for feature extraction and fast-to-train Kernel-based classifiers to adapt to the presence of novel objects or different domains. Additionally, a training protocol is proposed to shorten the training time.
IEEE TRANSACTIONS ON ROBOTICS
(2022)
Article
Computer Science, Artificial Intelligence
Qing Guo, Fazhi He, Bo Fan, Yupeng Song, Jicheng Dai, Linkun Fan
Summary: This paper proposes a neural framework named WalkFormer that applies a transformer to random walks in 3D meshes, enabling the utilization of semantic information. The framework incorporates a novel relative position encoding module and a parallelized execution method, improving computational efficiency. Experimental results demonstrate the effectiveness of the proposed method in typical 3D shape analysis tasks.
NEURAL COMPUTING & APPLICATIONS
(2023)
Article
Environmental Sciences
Donato Amitrano, Gerardo Di Martino, Antonio Iodice, Daniele Riccio, Giuseppe Ruello
Summary: This paper presents a new technique for mapping urban areas using multitemporal synthetic aperture radar data. The proposed methodology combines innovative RGB composites with self-organizing map (SOM) clustering and object-based image analysis. The technique has been tested in different scenarios in Italy and Germany, showing good agreement with the Urban Atlas of the European Environmental Agency.
Article
Computer Science, Artificial Intelligence
Rongkang Li, Yumeng Zhang, Dongmei Niu, Guangchao Yang, Numan Zafar, Caiming Zhang, Xiuyang Zhao
Summary: This paper introduces the usage of point convolution and point pooling on irregular point cloud data for learning high-level features. By applying these techniques, accurate shape information and geometric representation of point clouds can be achieved, leading to outstanding results in tasks like object classification and part segmentation.
Article
Environmental Sciences
Alexandru Pop, Victor Domsa, Levente Tamas
Summary: In this paper, a novel rotation normalization technique using an oriented bounding box for point cloud processing is proposed. It is used to create a point cloud annotation tool for part segmentation and trained on custom datasets for classification and part segmentation tasks. The method is successfully deployed on an embedded device with limited processing power and compared with other rotation-invariant features in noisy synthetic datasets. Our method offers more auxiliary information related to the object's dimension, position, and orientation while performing at a similar level.
Article
Computer Science, Artificial Intelligence
Guorong Li, Dexiang Hong, Kai Xu, Bineng Zhong, Li Su, Zhenjun Han, Qingming Huang
Summary: This paper proposes a self-supervised progressive network (SSPNet) for video object segmentation. SSPNet consists of a memory retrieval module (MRM) and a collaborative refinement module (CRM) to improve the performance. The MRM generates propagated coarse masks through self-supervised pixel-level and frame-level similarity learning, while the CRM refines the masks through cycle consistency region tracking. Novel mask-generation strategies are also designed to incorporate meaningful semantic information. Experimental results on multiple datasets demonstrate the superiority of SSPNet over state-of-the-art self-supervised methods, narrowing the gap with fully supervised methods.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Computer Science, Information Systems
Md Alamgir Hossain, Md Imtiaz Hossain, Md Delowar Hossain, Eui-Nam Huh
Summary: Real-time moving object detection is a crucial task in various fields, and accurately detecting objects in challenging backgrounds remains a major challenge. To address this, we propose a background subtraction-based method that dynamically combines different feature spaces using weighted fusion. Our method utilizes color-gradient background difference and segmentation noise to modify thresholds and background samples, achieving the best trade-off between accuracy and complexity compared to existing approaches.
MULTIMEDIA TOOLS AND APPLICATIONS
(2022)
Article
Engineering, Electrical & Electronic
Weide Liu, Guosheng Lin, Tianyi Zhang, Zichuan Liu
Summary: The study focuses on online semi-supervised video object segmentation and introduces a GCSeg network that achieves state-of-the-art performance by incorporating relationships at different time scales and an adaptive search strategy.
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY
(2021)
Article
Biochemical Research Methods
Sul-Hee Kim, Jin Kim, Su Yang, Sung-Hye Oh, Seung-Pyo Lee, Hoon Joo Yang, Tae-Il Kim, Won-Jin Yi
Summary: This study proposes a method for automatic segmentation of tooth enamel and alveolar bone using convolutional neural network (CNN) and quantitatively and automatically measuring the alveolar bone level (ABL) in optical coherence tomography (OCT) images. The experimental results demonstrate high segmentation accuracy in the tooth enamel and alveolar bone regions, and the measured results show a high correlation and reliability with the ground truth in OCT images.
BIOMEDICAL OPTICS EXPRESS
(2022)