检索结果-内蒙古大学图书馆

7th International Conference on computer science and Application Engineering, CSAE 2023

作者： Liang, Zhijia Zhang, Hui Cao, Yihong Du, Rui Biekezat, Baheti Huang, Zhihong School of Electrical and Information Engineering Changsha University of Science and Technology Changsha China School of Robotics Hunan University Changsha China College of Computer Science and Electronic Engineering Hunan University Changsha China State Grid Hunan Electric Power Company Limited Research Changsha China

ISBN: (纸本)9798400700590

Presence1 of foreign objects on power transmission lines can jeopardize the safe operation of the transmission lines. However, foreign objects on transmission lines are small in size and the available data for collection is limited. When using few-shot object detection algorithms for detection, there is a high risk of misclassifying of few-shot data and network bias, making the task of foreign object detection on transmission lines challenging. To address this, we propose a method for detecting foreign objects on transmission lines based on few-shot object detection. Firstly, we introduce an attention mechanism in transfer learning to enhance the network's focus on few-shot data and mitigate network bias. Secondly, we propose a method for online difficult sample selection to address the issues of few-shot data classification and the occurrence of false positives and false negatives due to small object sizes. Additionally, we used involution to improve the accuracy of foreign object detection. Furthermore, we construct the finely labeled Transmission Line Foreign Object Dataset (TLFOD). Our method achieves an average precision of 98.4% in foreign object detection on the challenging TLFOD dataset, surpassing existing state-of-the-art methods and demonstrating outstanding performance in classification regression and position regression. © 2023 ACM.

关键词： Object detection

来源：评论

学校读者我要写书评

暂无评论

MvEchoSeg: One-shot In-context Learning for Multi-view Echocardiography Segmentation

MvEchoSeg: One-shot In-context Learning for Multi-view Echoc...

引用

2024 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2024

作者： Guo, Yuting Duan, Ya Song, Wenfeng Li, Nannan Li, Aili Li, Shuai Beihang University State Key Laboratory of Virtual Reality Technology and Systems Beijing100191 China University of Chinese Academy of Sciences State Key Laboratory of Robotics Shenyang Institute of Automation Chinese Academy of Sciences Shenyang110016 China Beijing Information Science and Technology University School of Computer Science Beijing100192 China Dalian Maritime University School of Information Science and Technology Dalian116000 China China-Japan Friendship Hospital Beijing100029 China

ISBN: (纸本)9798350386226

Echocardiography is the clinical standard for evaluation of cardiac morphology, and function, and providing hemodynamic parameters in patients with known or suspected heart disease. Due to the diversity and complexity of diagnostic tasks, comprehensive interpretation of echocardiography often requires multi-view imaging for integrated metric analysis. Deep learning methods have become the mainstream approach for echocardiography segmentation. However, achieving segmentation of multi-view echocardiography still requires a substantial amount of annotation. To remedy this, we propose a one-shot in-context learning network mvEchoSeg for multi-view echocardiography segmentation. This network requires only one annotated image for each view. Specifically, we propose a Task Prompt Identifier (TPI) module to identify the task type of the image and allocate the most precise task prompt for it, with minimal adaptation of CLIP and few-shot strategies. Additionally, we leverage a unified In-Context Model (ICM) capable of performing a diverse set of echocardiography segmentation tasks automatically. Using fine-tuning and low-rank adapters improved the performance of the pre-trained model, achieving significant results with minimal training cost. Furthermore, we collect a multi-view echocardiography dataset (MVECD) with 8 views to evaluate our method. The results show an improvement of more than 10% in the DICE score compared to the SOTA foundational medical image segmentation models. To our knowledge, this is the first exploration of a one-shot model for multi-view echocardiography segmentation. Our codes and models are available at https://***/stellating/mvEchoSeg. © 2024 IEEE.

关键词： Image segmentation

来源：评论

学校读者我要写书评

暂无评论

A Research on Scattering Removal technology Using Compact GPU Machine for Real-Time Visualization 23

A Research on Scattering Removal Technology Using Compact GP...

引用

23rd International Conference on Control, Automation and Systems, ICCAS 2023

作者： Ono, Seiya Kim, Hyun-Woo Cho, Myungjin Lee, Min-Chul Kyushu Institute of Technology Department of Computer Science and Networks Fukuoka Iizuka820-8502 Japan School of ICT Robotics and Mechanical Engineering Iitc Hankyong National University Kyonggi-do Anseong17579 Korea Republic of

ISBN: (纸本)9788993215274

In this paper, we propose a system that enables visualization under the situation of scattering media such as fog and smoke by Peplography which is scattering media removal system, on a small GPU machine. Compared to conventional haze removal techniques, peplography technology is capable of removing more dense scattering media and can handle cases where the density of fog varies in the image. On the other hand, the conventional peplography process is performed on a desktop PC, and no system has yet been developed to perform peplography in an outside environment. Thus, we have developed a peplography system using the Jetson AGX Xavier, a single-board computer equipped with a GPU, which is superior to conventional methods in terms of portability, execution speed, and power consumption. We also demonstrate the performance of this system by conducting experiments using this system under a simulated scattering media scene and evaluating the obtained results in terms of FPS and the image evaluation metric. © 2023 ICROS.

关键词： Graphics processing unit

来源：评论

学校读者我要写书评

暂无评论

Analysis of Accurate Focal Points and Automatic Method of Digital Holographic Microscopy (DHM) for Obtaining an Accurate 3D Profile 23

Analysis of Accurate Focal Points and Automatic Method of Di...

引用

23rd International Conference on Control, Automation and Systems, ICCAS 2023

作者： Ono, Taishi Kim, Hyun-Woo Cho, Myungjin Lee, Min-Chul Kyushu Institute of Technology Department of Computer Science and Networks Fukuoka Iizuka820-8502 Japan School of ICT Robotics and Mechanical Engineering Iitc Hankyong National University Kyonggi-do Anseong17579 Korea Republic of

ISBN: (纸本)9788993215274

Currently, Digital Holographic Microscopy(DHM) is used for researching disease diagnosis or microbes. It cannot obtain the correct three-dimensional (3D) profile for a little noise because biological cells are microscopic. Then we need images of the focal point of DHM for reducing noise. However, in DHM, the depth of field becomes smaller influenced by the objective lens. In addition, focusing to obtain hologram images of biological cells in conventional DHM systems is entirely manual, because the decision of the focal point is made by the photographer's senses, which makes it even more difficult to focus on the object. These factors make it difficult for the conventional DHM system to acquire images of the focal points. Therefore, in this paper, we propose the analysis of accurate focusing points using a noise map. This eliminates the illusion or misperception that is a drawback of the manual. We improved an accurate 3D profile. 3D profiles were created at the focal point by the proposed method and at the positions before and after the focal points, and evaluated for SNR. From this, the proposed method can be used to analyze biological cells and microstructures. © 2023 ICROS.

关键词： Holograms

来源：评论

学校读者我要写书评

暂无评论

Improvement of Depth Mapping and Occlusion Removal Method Using Integral Imaging 23

Improvement of Depth Mapping and Occlusion Removal Method Us...

引用

23rd International Conference on Control, Automation and Systems, ICCAS 2023

作者： Sakanoue, Haruka Kim, Hyun-Woo Cho, Myungjin Lee, Min-Chul Kyushu Institute of Technology Department of Computer Science and Networks Fukuoka Iizuka820-8502 Japan School of ICT Robotics and Mechanical Engineering Iitc Hankyong National University Kyonggi-do Anseong17579 Korea Republic of

ISBN: (纸本)9788993215274

In this paper, we propose the depth mapping and occlusion removal method using integral imaging (InIm). InIm is a method to obtain 3D information by capturing and reconstructing an object from multiple viewpoints. InIm can remove occlusion using parallax obtained by photographing an object from multiple viewpoints. However, occlusion removal uses a unit matrix to calculate the number of overlaps, which results in an overall darkening of the image. To solve this problem, this paper aims to improve occlusion removal by using the matrix of the mask image to calculate the number of overlaps. InIm can provide depth mapping that shows the distance from an object as pixel values. However, when there is an occlusion in front of the object, it is difficult to provide depth mapping object due to the influence of the occlusion. To solve this problem, we propose a method of depth mapping and occlusion removal method using InIm. Our method is expected to make great progress in improving safety in the field of autonomous driving. © 2023 ICROS.

关键词： Mapping

来源：评论

学校读者我要写书评

暂无评论

Photon-Counting Integral Imaging Using Bayesian Estimation with Depth Camera Information 23

Photon-Counting Integral Imaging Using Bayesian Estimation w...

引用

23rd International Conference on Control, Automation and Systems, ICCAS 2023

作者： Uchino, Hideaki Kim, Hyun-Woo Cho, Myungjin Lee, Min-Chul Kyushu Institute of Technology Department of Computer Science and Networks Fukuoka Iizuka820-8502 Japan School of ICT Robotics and Mechanical Engineering Iitc Hankyong National University Kyonggi-do Anseong17579 Korea Republic of

ISBN: (纸本)9788993215274

In this paper, we propose a method for visualizing object information under low light conditions using photon-counting integral imaging and depth images as prior information. To visualize 3D objects under low-light conditions, computational photon-counting imaging may be used. It is possible to estimate photons using statistical methods such as the Poisson distribution. Therefore, photon-counting integral imaging can visualize 3D images under low-light conditions by maximum likelihood estimation (MLE). However, MLE does not consider prior information so it may not be an accurate estimation. In addition, it is difficult to obtain accurate depth information as the distance increases due to rounding errors caused by computational reconstruction. Therefore, we propose Bayesian estimation such as a maximum a posteriori estimation method that uses as a priori information images from a depth camera, which can be acquired even in low-light conditions, to improve the quality of 3D images. To evaluate image quality of the proposed method, optical experiments were conducted by calculating image quality metrics. This technology can be utilized for night vision applications, such as object recognition in security camera systems and autonomous driving, using RGB cameras and depth cameras. © 2023 ICROS.

关键词： Maximum likelihood estimation

来源：评论

学校读者我要写书评

暂无评论

A Study on Coherence Length and Interferometry in Digital Holographic Microscopy 23

A Study on Coherence Length and Interferometry in Digital Ho...

引用

23rd International Conference on Control, Automation and Systems, ICCAS 2023

作者： Koga, Takahiro Kim, Hyun-Woo Cho, Myungjin Lee, Min-Chul Kyushu Institute of Technology Department of Computer Science and Networks Fukuoka Iizuka820-8502 Japan School of ICT Robotics and Mechanical Engineering Iitc Hankyong National University Kyonggi-do Anseong17579 Korea Republic of

ISBN: (纸本)9788993215274

Digital Holographic Microscopy (DHM) can display 3D profiles of microscopic objects and is expected to be used in a wide range of fields such as microbiology, microstructural investigation, and disease diagnosis. Since these fields require detailed information of the object, acquiring an accurate 3D profile is the most important. However, DHM requires several experimental equipments, and it does not define appropriate lasers for interferometers. Therefore, in this paper, we implement an experiment and find the optimum laser for the most accurate 3D profiles of interferometers by comparing 3D profiles using two types of interferometers and three types of lasers with different wavelengths. The results show that the relationship between the laser's coherence length and the optical path length of the interferometer is related to the accuracy of the 3D profiles. They also show that the accuracy of the 3D profile is higher when the coherence length is longer than the optical path length plus twice the height of the object. © 2023 ICROS.

关键词： Computational Imaging

来源：评论

学校读者我要写书评

暂无评论

RetouchGPT: LLM-based Interactive High-Fidelity Face Retouching via Imperfection Prompting 39

RetouchGPT: LLM-based Interactive High-Fidelity Face Retouch...

引用

39th Annual AAAI Conference on Artificial Intelligence, AAAI 2025

作者： Xue, Wen Ding, Chun Xu, Ruotao Wu, Si Xu, Yong Wong, Hau-San School of Computer Science and Engineering South China University of Technology China Institute of Super Robotics (Huangpu) China Department of Computer Science City University of Hong Kong Hong Kong

ISBN: (纸本)157735897X

Face retouching aims to remove facial imperfections from image and videos while at the same time preserving face attributes. The existing methods are designed to perform non-interactive end-to-end retouching, while the ability to interact with users is highly demanded in downstream applications. In this paper, we propose RetouchGPT, a novel framework that leverages Large Language Models (LLMs) to guide the interactive retouching process. Towards this end, we design an instruction-driven imperfection prediction module to accurately identify imperfections by integrating textual and visual features. To learn imperfection prompts, we further incorporate a LLM-based embedding module to fuse multi-modal conditioning information. The prompt-based feature modification is performed in each transformer block, such that the imperfection features are suppressed and replaced with the features of normal skin progressively. Extensive experiments have been performed to verify effectiveness of our design elements and demonstrate that RetouchGPT is a useful tool for interactive face retouching and achieves superior performance over state-of-the-arts. Copyright © 2025, Association for the Advancement of Artificial Intelligence (***). All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Noise Removal in Scattering Media Enviorment Using Peploaraphy and DCLGAN 23

Noise Removal in Scattering Media Enviorment Using Peploarap...

引用

23rd International Conference on Control, Automation and Systems, ICCAS 2023

作者： Numata, Riki Kim, Hyun-Woo Ono, Seiya Cho, Myungjin Lee, Min-Chul Kyushu Institute of Technology Department of Computer Science and Networks Fukuoka Iizuka820-8502 Japan School of ICT Robotics and Mechanical Engineering Iitc Hankyong National University Kyonggi-do Anseong17579 Korea Republic of

ISBN: (纸本)9788993215274

Noise removal in scattering media environments such as fog and smoke are an important task in reality. For example, it can be applied in many fields such as searching under turbid water and rescuing in smoke. Therefore, the purpose of this research is to remove the noise under the scattering medium environment and to visualize the object. As a conventional method for removing noise, there is a technique called peplography. Peplograhy enables the denoising of dense fog and reconstructing the image in scattering media environments. However, peplography has a problem that complex noise exists after removing particularly deep noise, and as a result, it becomes difficult to visualize the object. This research aims to solve this problem. We propose a new denoising technology that combines peplography with DCLGAN. DCLGAN is an image transformation algorithm using deep learning and an improved model of CycleGAN which is also used in the field of noise reduction. In this study, we evaluate the conventional method and the proposed method for scattering media images. As the evaluation method, visual evaluation and numerical evaluation using the PSNR index are performed to demonstrate the effectiveness of the proposed method. We believe that the proposed method can be applied in various fields such as object visualization under turbid water and automatic driving technology with fog removed. © 2023 ICROS.

关键词： Smoke

来源：评论

学校读者我要写书评

暂无评论

MQADet: A Plug-and-Play Paradigm for Enhancing Open-Vocabulary Object Detection via Multimodal Question Answering

arXiv

引用

arXiv 2025年

作者： Li, Caixiong Zhao, Xiongwei Zhang, Jinhang Zhang, Xing Sun, Qihao Wu, Zhou School of Computer and Information Science Qinghai Institute of Technology China School of Information Science and Technology Harbin Institute of Technology Shen Zhen China The State Key Laboratory of Robotics and System Harbin Institute of Technology China School of Computer Science and Technology Qinghai University China Eryuan Digital Technology Co. Ltd. China

Open-vocabulary detection (OVD) is a challenging task to detect and classify objects from an unrestricted set of categories, including those unseen during training. Existing open-vocabulary detectors are limited by complex visual-textual misalignment and long-tailed category imbalances, leading to suboptimal performance in challenging scenarios. To address these limitations, we introduce MQADet, a universal paradigm for enhancing existing open-vocabulary detectors by leveraging the cross-modal reasoning capabilities of multimodal large language models (MLLMs). MQADet functions as a plug-and-play solution that integrates seamlessly with pre-trained object detectors without substantial additional training costs. Specifically, we design a novel three-stage Multimodal Question Answering (MQA) pipeline to guide the MLLMs to precisely localize complex textual and visual targets while effectively enhancing the focus of existing object detectors on relevant objects. To validate our approach, we present a new benchmark for evaluating our paradigm on four challenging open-vocabulary datasets, employing three state-of-the-art object detectors as baselines. Experimental results demonstrate that our proposed paradigm significantly improves the performance of existing detectors, particularly in unseen complex categories, across diverse and challenging scenarios. To facilitate future research, we will publicly release our code. © 2025, CC BY.

关键词： Object detection

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：