Article
Computer Science, Information Systems
Sheela Ramachandra, Suchithra Ramachandran
Summary: This paper proposes a Periocular recognition algorithm that utilizes region-specific and sub-image-based neighbor gradient feature extraction to achieve better recognition results. The proposed method segments the periocular region into sub-regions and extracts features using different algorithms. Experimental results demonstrate that the proposed method outperforms traditional algorithms on multiple datasets.
JOURNAL OF KING SAUD UNIVERSITY-COMPUTER AND INFORMATION SCIENCES
(2022)
Article
Robotics
Kuan Xu, Chen Wang, Chao Chen, Wei Wu, Sebastian Scherer
Summary: Object encoding and identification are crucial for robotic tasks. This letter proposes a novel object encoding method called AirCode based on a graph of key-points. It achieves robustness to viewpoint changes, scaling, occlusion, and object deformation through feature sparse encoding and object dense encoding. Experimental results show that it outperforms state-of-the-art algorithms in object identification and provides reliable semantic relocalization.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2022)
Article
Computer Science, Artificial Intelligence
Zhitong Xiong, Yuan Yuan, Qi Wang
Summary: An efficient framework for RGB-D scene recognition is proposed in this article, which adaptively selects important local features to capture the spatial variability of scene images. By designing a differentiable local feature selection (DLFS) module, key local scene-related features can be extracted from spatially-correlated multi-modal RGB-D features. By concatenating local-orderless and global-structured multi-modal features, the proposed framework achieves state-of-the-art performance on public RGB-D scene recognition datasets.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2021)
Article
Computer Science, Artificial Intelligence
Zhiliang Peng, Zonghao Guo, Wei Huang, Yaowei Wang, Lingxi Xie, Jianbin Jiao, Qi Tian, Qixiang Ye
Summary: This paper proposes a hybrid network structure called Conformer, which combines the advantages of convolution operations and self-attention mechanisms for enhanced representation learning.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Purandhar Reddy Vadamala, Annis Fathima Aklak
Summary: In this paper, a visual object tracking framework based on object appearance feature update is proposed. The tracking model utilizes template spatial information and object features to track the target object in successive frames, and adapts to appearance variations by updating the tracked template feature vector. Experimental results demonstrate its high precision (84.5%) and tracking speed (10.1 fps), outperforming conventional trackers in qualitative analysis.
Article
Computer Science, Information Systems
Alaa Eleyan
Summary: This study investigates the impact of feature fusion on face recognition performance by fusing different feature descriptors. The results show that fused feature descriptors can significantly improve performance, especially when the training set is limited.
JOURNAL OF KING SAUD UNIVERSITY-COMPUTER AND INFORMATION SCIENCES
(2023)
Article
Computer Science, Information Systems
Sunyoung Cho, Jwajin Lee
Summary: Facial expression recognition (FER) is a challenging task, especially under unconstrained conditions with variant head poses. To address this problem, the authors propose a local attention network (LAN) that adaptively captures important facial regions based on pose variations. The LAN improves FER performance by emphasizing attentive regions and suppressing regions that are not differentiated between classes. Experiments on multiple datasets demonstrate the effectiveness of the LAN and its superiority compared to previous methods.
Article
Computer Science, Artificial Intelligence
Charulata Patil, Aditya Abhyankar
Summary: Scene graphs present the semantic highlights of the underlying image in directed graph form. Their automated generation often neglects the diverse attributes of the objects. We propose a Multiple Attribute Detector that can capture structured attribute information of an object, including attribute type and value. This module can generate multiple triplets for each detected object in the scene and be integrated with existing scene graph generation frameworks.
MACHINE VISION AND APPLICATIONS
(2023)
Article
Automation & Control Systems
Hongsen Liu, Yang Cong, Gan Sun, Yandong Tang
Summary: The study proposes a hand-crafted feature descriptor called VSLPS for 3-D object recognition, which solves the feature-points matching problem through a voting strategy, and experiments show its excellent performance.
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS
(2021)
Article
Computer Science, Artificial Intelligence
Xiaojun Chang, Pengzhen Ren, Pengfei Xu, Zhihui Li, Xiaojiang Chen, Alex Hauptmann
Summary: Scene graph is a structured representation of a scene, expressing objects, attributes, and relationships. With the development of computer vision, people aim for a higher level of understanding and reasoning about visual scenes. Scene graphs have attracted researchers' attention as a powerful tool for scene understanding.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2023)
Article
Environmental Sciences
Bin Lin, Yunpeng Bai, Bendu Bai, Ying Li
Summary: This paper proposes a robust CF-based tracker with feature integration and response map enhancement to address the challenges in selecting suitable features and alleviating model drift for online UAV tracking. Experiments show that the proposed tracker outperforms other algorithms, achieving real-time tracking speed and efficient application in UAV tracking scenarios.
Article
Computer Science, Artificial Intelligence
Ningyu Zhang, Shumin Deng, Hongbin Ye, Wei Zhang, Huajun Chen
Summary: Recent research has shown that previous approaches may fail in handling text ambiguities in similar contexts and lead to contradictions in commonsense. Inspired by capsule networks and implicit entity-relation schema, a novel Cascade Bidirectional Capsule Network is proposed to address these issues. The experimental results demonstrate that the proposed approach is more efficient and has a stronger generalization ability to handle complex surface forms.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Computer Science, Information Systems
Daihui Li, Feng Liu, Tongsheng Shen, Liang Chen, Dexin Zhao
Summary: In this paper, a robust feature extraction method based on multi-task learning is proposed to solve the complex problems of target classification and recognition in underwater acoustic signal processing. The method optimizes feature extraction by deploying a multi-task learning model and improves the robustness and representation of classification features. Experimental results show that the proposed method effectively improves recognition accuracy and maintains high performance under different noise levels.
Article
Geochemistry & Geophysics
Jun Chen, Meng Yang, Chengli Peng, Linbo Luo, Wenping Gong
Summary: This article proposes a simple and effective feature matching method for remote sensing image registration. By using local consensus constraints to remove outliers, the method can solve the problem of high outliers ratio caused by nonrigid transformation, nonlinear radiation difference, and speckle noise. Experiments on remote sensing images demonstrate the superiority of this method in feature matching.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Computer Science, Artificial Intelligence
Shaochuan Zhao, Tianyang Xu, Xiao-Jun Wu, Josef Kittler
Summary: This research proposes a novel visual object tracking method that improves the robustness and temporal stability of the tracker by using a cross channel correlation mechanism and a jitter metric.
INTERNATIONAL JOURNAL OF COMPUTER VISION
(2023)
Article
Computer Science, Artificial Intelligence
Han Chen, Yifan Jiang, Murray Loew, Hanseok Ko
Summary: This paper presents an unsupervised domain adaptation based segmentation network method, which combines synthetic data and limited unlabeled real data to improve the segmentation performance of infection areas in COVID-19 CT images. Experimental results demonstrate that this method achieves state-of-the-art segmentation performance on COVID-19 CT images.
APPLIED INTELLIGENCE
(2022)
Article
Geochemistry & Geophysics
Shou Zhang, Bonhwa Ku, Hanseok Ko
Summary: Recently, research has focused on establishing an early warning system for earthquakes by analyzing short seismic waves to minimize damage. This letter proposes an improved ConvNetQuake method for earthquake event classification by adding learnable features related to the maximum amplitude of seismic waveform, achieving significant performance improvement on earthquake event classification.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Geochemistry & Geophysics
Bokyeung Lee, Bonhwa Ku, Wanjin Kim, Seungil Kim, Hanseok Ko
Summary: In this study, a learning-based compressive sensing algorithm is proposed for denoising sonar images. The method combines deep learning and iterative shrinkage and thresholding algorithm, and incorporates CoordConv to help remove nonhomogeneous noise. Experimental results show that the proposed method outperforms existing methods in terms of noise removal and memory requirements.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Geochemistry & Geophysics
Gwantae Kim, Bonhwa Ku, Jae-Kwang Ahn, Hanseok Ko
Summary: This study proposes a multiple station-based seismic event classification model using a deep convolution neural network and graph convolution network. The model shows superior performance in classifying various seismic events and reduces false alarms when using continuous waveforms.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Geochemistry & Geophysics
Jeongki Min, Bonwha Ku, Hanseok Ko
Summary: The letter proposes an earthquake event classification model utilizing a feedback network and curriculum learning, effectively using feature concatenation and gated convolution for CL. Through comparison experiments with existing models on Korean Peninsula and Stanford earthquake datasets, the effectiveness of the proposed model is demonstrated.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Computer Science, Information Systems
Heewoong Ahn, Sunhwa Lee, Hanseok Ko, Meejoung Kim, Sung Won Han, Junhee Seok
Summary: A convolutional autoencoder model is proposed to help weather forecasters analyze current weather status by finding similar weather maps from the past based on latent features extraction. Similarity between images is measured using metrics like mean squared error and structural similarity, and case studies are conducted to visualize the results. The paper demonstrates the usefulness of searching similar weather maps for all forecasters in analyzing and predicting the weather.
Article
Engineering, Biomedical
Han Chen, Yifan Jiang, Hanseok Ko, Murray Loew
Summary: This paper proposes a novel unsupervised method to improve the generalization ability of the segmentation network for COVID-19 CT images by learning domain-invariant features from lung cancer and COVID-19 images.
BIOMEDICAL SIGNAL PROCESSING AND CONTROL
(2023)
Article
Geochemistry & Geophysics
Dongsik Yoon, Yuanming Li, Bonhwa Ku, Hanseok Ko
Summary: Estimating earthquake parameters is crucial for an earthquake analysis system. This letter proposes a novel estimation method using multitasking deep learning and a convolutional recurrent neural network (CRNN) with only a single station. The method accurately estimates earthquake magnitude using the stream maximum of the input waveform. The high performance of the proposed method is verified through evaluation using the Stanford Earthquake dataset (STEAD) and the Kiban Kyoshin Network (KiK-net) dataset.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2023)
Article
Computer Science, Artificial Intelligence
Bokyeung Lee, Kyungdeuk Ko, Jonghwan Hong, Hanseok Ko
Summary: Driven by the learning ability of deep networks, generalized models have been proposed for single-image super-resolution tasks using external datasets. However, models trained solely on external data may struggle to super-resolve images in unfamiliar domains. To address this limitation, internal learning approaches have been proposed, but they often suffer from poor performance due to a fixed architecture. Therefore, we propose a novel training process that combines external and internal learning, allowing the network to adapt to specific test images. Our approach outperforms state-of-the-art super-resolution algorithms in various image domains and produces impressive results in terms of texture and color accuracy.
Article
Computer Science, Information Systems
Jonghwan Hong, Bokyeung Lee, Kyungdeuk Ko, Hanseok Ko
Summary: Although traditional methods have achieved significant performance improvement for image super-resolution, their high computational cost limits their real-world application. In this paper, a fast non-local attention network (FNLNET) is proposed for super light image super-resolution, which can capture global representation.
JOURNAL OF VISUAL COMMUNICATION AND IMAGE REPRESENTATION
(2023)
Article
Engineering, Electrical & Electronic
Bokyeung Lee, Donghyeon Kim, Gwantae Kim, Hanseok Ko
Summary: This letter proposes Channel Shuffle Neural Architecture Search (CSNAS) with channel weights for Key-Word Spotting (KWS). CSNAS can simultaneously control the number of parameters, FLOPs, and performance in the search process, and generate network architectures that outperform state-of-the-art KWS methods.
IEEE SIGNAL PROCESSING LETTERS
(2023)
Article
Engineering, Electrical & Electronic
Donghyeon Kim, Kyungdeuk Ko, David K. Han, Hanseok Ko
Summary: Keyword Spotting is crucial for smart devices to respond to user commands, and the LOVO loss introduced in this study helps enhance the network's ability to extract discriminative features in noisy environments.
IEEE SIGNAL PROCESSING LETTERS
(2022)
Article
Engineering, Electrical & Electronic
Bokyeung Lee, Kyungdeuk Ko, Jonghwan Hong, Bonhwa Ku, Hanseok Ko
Summary: This paper proposes a novel training process that simultaneously learns the sensing and decoder networks using Information Bottleneck theory, to improve the reconstruction performance of image compressed sensing algorithms.
IEEE SIGNAL PROCESSING LETTERS
(2022)
Proceedings Paper
Computer Science, Interdisciplinary Applications
Ange Lou, Shuyue Guan, Hanseok Ko, Murray Loew
Summary: Accurate segmentation of medical images is crucial for disease diagnosis and treatment. Existing networks often have poor performance in segmenting small objects. This paper proposes CaraNet, which shows distinct advantages in segmenting small medical objects.
MEDICAL IMAGING 2022: IMAGE PROCESSING
(2022)
Article
Computer Science, Information Systems
Han Chen, Yifan Jiang, Hanseok Ko
Summary: PG-GCN is a multi-modal framework that explores robust features from both pose and skeleton data simultaneously, with early-stage feature fusion using a dynamic attention module, achieving state-of-the-art performance in action recognition tasks.