检索结果-内蒙古大学图书馆

Imaginations Generate Images for Multi-modal Machine Translation 1

13th International Conference on Computer Engineering and Networks (CENet)

作者： Yang, Xiaona Sun, Wenli Wei, Wei Li, Yinlin Shi, Xiayang Zhengzhou Univ Light Ind Zhengzhou Peoples R China Wuhan Second Ship Design & Res Inst Wuhan Peoples R China Chinese Acad Sci Inst Automat State Key Lab Management & Control Complex Syst Beijing 100190 Peoples R China

ISBN: (数字)9789819992430

ISBN: (纸本)9789819992454;9789819992430;9789819992423

Multi-modal machine translation (MMT) aims at exploring better translation systems by integrating the visual annotation which presents the content described in the bilingual parallel sentence pair into the conventional only-text neural machine translation (NMT). However, existing methods heavily rely on the manual annotated images data set. The cost of manual image annotation is relatively high at this stage. In this paper, we propose the generative imagination network with transformer to automatically generate visual annotations semantic-equivalent with source and target sentences. The proposed model receives the inputs of source-target bilingual sentences and generates visual annotations for MMT. Experiments analysis demonstrate that our model can generate high-quality annotated images and prompt the performance of MMT. Additionally, we use our model to generate annotated images for a famous English-German IWSLT-2015, the experimental results show the improvement for MMT.

关键词： Multi-modal machine translation Visual annotation

来源：评论

学校读者我要写书评

暂无评论

LC-LLM: Explainable lane-change intention and trajectory predictions with Large Language Models

引用

COMMUNICATIONS IN TRANSPORTATION RESEARCH 2025年 5卷

作者： Peng, Mingxing Guo, Xusen Chen, Xianda Chen, Kehua Zhu, Meixin Chen, Long Wang, Fei-Yue Hong Kong Univ Sci & Technol Guangzhou Intelligent Transportat INTR Syst Hub Guangzhou 511400 Peoples R China Hong Kong Univ Sci & Technol Acad Interdisciplinary Studies Div Emerging Interdisciplinary Areas EMIA Hong Kong 999077 Peoples R China Guangdong Prov Key Lab Integrated Commun Sensing & Guangzhou 511400 Peoples R China Chinese Acad Sci Inst Automat State Key Lab Management & Control Complex Syst Beijing 100190 Peoples R China

To ensure safe driving in dynamic environments, autonomous vehicles should possess the capability to accurately predict lane change intentions of surrounding vehicles in advance and forecast their future trajectories. Existing motion prediction approaches have ample room for improvement, particularly in terms of long-term prediction accuracy and interpretability. In this study, we address these challenges by proposing a Lane Change-Large Language Model (LC-LLM), an explainable lane change prediction model that leverages the strong reasoning capabilities and self explanation abilities of Large Language Models (LLMs). Essentially, we reformulate the lane change prediction task as a language modeling problem, processing heterogeneous driving scenario information as natural language prompts for LLMs and employing supervised fine-tuning to tailor LLMs specifically for lane change prediction task. Additionally, we finetune the Chain-of-Thought (CoT) reasoning to improve prediction transparency and reliability, and include explanatory requirements in the prompts during the inference stage. Therefore, our LC-LLM not only predicts lane change intentions and trajectories but also provides CoT reasoning and explanations for its predictions, enhancing its interpretability. Extensive experiments based on the large-scale highD dataset demonstrate the superior performance and interpretability of our LC-LLM in lane change prediction task. To the best of our knowledge, this is the first attempt to utilize LLMs for predicting lane change behavior. Our study shows that LLMs can effectively encode comprehensive interaction information for understanding driving behavior.

关键词： Lane change (LC) Large language models (LLMs) Intention prediction Trajectory prediction Fine-tuning Interpretability Autonomous driving

来源：评论

学校读者我要写书评

暂无评论

Sustainability Opportunities and Ethical Challenges of AI-Enabled Connected Autonomous Vehicles Routing in Urban Areas

IEEE TRANSACTIONS ON INTELLIGENT VEHICLES

引用

IEEE TRANSACTIONS ON INTELLIGENT VEHICLES 2024年第1期9卷 55-58页

作者： Guo, Rongge Vallati, Mauro Wang, Yutong Zhang, Hui Chen, Yuanyuan Wang, Fei-Yue Univ Huddersfield Sch Comp & Engn Huddersfield HD13DH England Chinese Acad Sci Inst Automat State Key Lab Management & Control Complex Syst Beijing 100190 Peoples R China Beihang Univ Sch Transportat Sci & Engn Beijing 100091 Peoples R China Chinese Acad Sci Inst Automat State Key Lab Multimodal Artificial Intelligence Beijing 100190 Peoples R China Macau Univ Sci & Technol Inst Engn Macau 999078 Peoples R China

The advent of Connected Autonomous Vehicles (CAVs) paves the way to a new era of urban traffic control and management, driven by Artificial Intelligence (AI)-enabled strategies. This advancement promises significant improvements in infrastructure use optimization, traffic delay reduction, and overall sustainability. The autonomous driving capabilities of CAVs, coupled with the communication technology, allow vehicles to play an active role in urban traffic control: they can follow tailored instructions and can act as highly accurate moving sensors for traffic authorities. However, such improved capabilities come at the cost of unprecedented vulnerabilities to cyber exploitation, and with the concrete potential to increase social and economic disparities. As an extension of TIV-DHW (Distributed/Decentralized Hybrid Workshop) on ERS (Ethics, Responsibility, and Sustainability), this letter explores how the AI-enabled routing methodology can enhance urban transportation sustainability while also discussing its ethical implications and challenges.

关键词： Routing Artificial intelligence Traffic control Sustainable development Urban areas Safety Ethics Connected autonomous vehicles artificial intelligence traffic routing sustainability ethics

来源：评论

学校读者我要写书评

暂无评论

CSIR: Cascaded Sliding CVAEs With Iterative Socially-Aware Rethinking for Trajectory Prediction

引用

IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION systemS 2023年第12期24卷 14957-14969页

作者： Zhou, Hao Yang, Xu Ren, Dongchun Huang, Hai Fan, Mingyu Harbin Engn Univ Natl Key Lab Sci & Technol Underwater Vehicle Harbin 150001 Peoples R China Chinese Acad Sci Inst Automat State Key Lab Management & Control Complex Syst Beijing 100190 Peoples R China Meituan Res Ctr Autonomous Vehicles Beijing 100102 Peoples R China Donghua Univ Inst Artificial Intelligence Shanghai 200051 Peoples R China

Pedestrian trajectory prediction is a hot research topic in many applications, such as video surveillance and autonomous driving. Although many efforts have been done on this topic, there are still many challenges, including accumulated prediction errors, insufficient training data usage, and future-past incompatibility. To overcome these challenges, we propose a novel trajectory prediction method, called CSIR, which consists of a cascaded sliding conditional variational autoencoder (CS-CVAE) module and an iterative future-past social compatible rethinking (I-SCR) module. The CS-CVAE module reduces the accumulated prediction errors by using cascaded prediction models for the early future time steps. In this way, the training losses of the early time steps are separately considered and minimized from the later losses. For the following time steps in CS-CVAE, a sliding prediction model with a longer observation time span is used and additional data from the future time span can be collected for training. On the other hand, the I-SCR module generates offsets to improve the predictions iteratively by checking the interaction compatibility between the predicted trajectories and the past trajectories, which resembles with the human rethinking mechanism in motion planning. Experiments results on two widely explored pedestrian trajectory prediction datasets, Stanford Drone Dataset (SDD) and ETH/UCY, show that the proposed method surpasses previous state-of-the-art methods by notable margins.

关键词： Cascaded prediction sliding sequence prediction iterative social-aware rethinking trajectory prediction

来源：评论

学校读者我要写书评

暂无评论

Advanced Scenario Generation for Calibration and Verification of Autonomous Vehicles

IEEE TRANSACTIONS ON INTELLIGENT VEHICLES

引用

IEEE TRANSACTIONS ON INTELLIGENT VEHICLES 2023年第5期8卷 3211-3216页

作者： Li, Xuan Teng, Siyu Liu, Bingzi Dai, Xingyuan Na, Xiaoxiang Wang, Fei-Yue Peng Cheng Lab Shenzhen 518000 Peoples R China Hong Kong Baptist Univ Hong Kong 999077 Peoples R China Chinese Acad Sci Inst Automat State Key Lab Management & Control Complex Syst Beijing 100190 Peoples R China Univ Cambridge Dept Engn Cambridge CB2 1PZ England

As driving scenarios and autonomous vehicles (AVs) become increasingly intricating, there is an increasing need for innovative frameworks that can enhance and test AV capabilities across diverse scenarios. At present, the design and validation for AVs predominantly rely on simulation software or real-road testing methodologies. However, these approaches possess inherent limitations, leading to inaccuracy, reduced efficiency, and potential hazards during the development process. This letter reports our first DHW (decentralized and hybrid workshop) on Scenarios Engineering (SE), that aims to calibrate and validate AV modules through advanced scenarios. Specifically, the DHW discusses combining SE with existing simulation software and real-road testing strategies, enabling developers to maximize the advantages of each approach while ensuring the safety and efficacy in the development of autonomous vehicles. Our findings demonstrate the benefits of enhancing realism of simulation scenarios in the aspects of both content and appearance. Furthermore, the use of scenarios engineering is explored in order to enhance the diversity and safety of real-road testing by means of integrating virtual and physical components.

关键词： Testing Safety Autonomous vehicles Roads ISO Standards Calibration Standards Scenarios engineering autonomous vehicles parallel intelligence simulation scenarios real-road testing

来源：评论

学校读者我要写书评

暂无评论

ResDO-UNet: A deep residual network for accurate retinal vessel segmentation from fundus images

引用

BIOMEDICAL SIGNAL PROCESSING AND control 2023年第Part1期79卷 000-000页

作者： Liu, Yanhong Shen, Ji Yang, Lei Bian, Guibin Yu, Hongnian Zhengzhou Univ Sch Elect & Informat Engn Zhengzhou 450001 Henan Peoples R China Robot Percept & Control Engn Lab Henan Prov Zhengzhou 450001 Peoples R China Chinese Acad Sci Inst Automat State Key Lab Management & Control Complex Syst Beijing 100190 Peoples R China Edinburgh Napier Univ Built Environm Edinburgh EH10 5DT Scotland

For the clinical diagnosis, it is essential to obtain accurate morphology data of retinal blood vessels from patients, and the morphology of retinal blood vessels can well help doctors to judge the patient's condition and give targeted therapeutic measures. Conventional manual retinal blood vessel segmentation by the doctors from the fundus images is time-consuming and laborious, while it also requires the rich doctor's expertise. With the strong context feature expression ability of deep convolutional neural networks (DCNN), it has shown a promising performance on retinal blood vessel segmentation, specially U-shape network (U-Net) and its variant. However, due to the information loss issue caused by multiple pooling operations and insufficient process issue of local context features by skip connections, most of segmentation methods still exist a certain shortcoming on accurate fine vessel detection. To address this issue, based on the encoder-decoder framework, a novel retinal vessel segmentation network, called ResDO-UNet, is proposed to provide an automatic and end-to-end detection scheme from fundus images. To enhance feature extraction capabilities, combined with depth-wise over-parameterized convolutional layer (DO-conv), a residual DO-conv (ResDO-conv) network is proposed to act as the backbone network to acquire strong context features. In addition, to reduce the effect of information loss caused by multiple pooling operations, taking advantages of max pooling and average pooling layers, a pooling fusion block (PFB) is proposed to realize nonlinear fusion pooling. Meanwhile, faced with insufficient process of local context features by skip connections, an attention fusion block (AFB) is proposed to realize effective multi-scale feature expression. Combined with the three public available data sets on retinal vessel segmentation, including DRIVE, STARE and CHASE_DB1, the proposed segmentation network could reach a state-of-the-art detection performance compa

关键词： Retinal vessels segmentation Deep learning U-Net network Residual DO-conv network

来源：评论

学校读者我要写书评

暂无评论

Embed Trajectory Imitation in Reinforcement Learning: A Hybrid Method for Autonomous Vehicle Planning 3

Embed Trajectory Imitation in Reinforcement Learning: A Hybr...

引用

3rd IEEE International Conference on Digital Twins and Parallel Intelligence, DTPI 2023

作者： Wang, Yuxiao Dai, Xingyuan Wang, Kara Ali, Hub Zhu, Fenghua Chinese Academy of Sciences State Key Laboratory of Management and Control for Complex System Institute of Automation Beijing China University of Chinese Academy of Sciences School of Artificial Intelligence Beijing China Anhui University School of Artificial Intelligence Hefei China

ISBN: (纸本)9798350318470

Learning-based autonomous vehicle trajectory planning methods have shown excellent performance in a variety of complex traffic scenarios. However, the existing imitation learning (IL) and reinforcement learning (RL) algorithms still have their limitations, such as poor safety and generalizability for IL, and low data efficiency for RL. To leverage their respective advantages and mitigate the limitations, this paper proposes a novel hybrid RL algorithm for autonomous vehicle planning, where IL is embedded in it to guide its exploration with expert knowledge. Different from existing approaches, we use multi-step trajectory prediction instead of behavior cloning as the IL method integrated with online RL. Through such design, we make a further step in the research about how expert demonstration can be helpful to RL. Moreover, we conduct parallel training and testing of the algorithm based on real-world driving data. Experimental results demonstrate that our proposed approach outperforms standalone IL and RL methods, and performs better than RL methods enhanced by behavior cloning. © 2023 IEEE.

关键词： Autonomous vehicles

来源：评论

学校读者我要写书评

暂无评论

Valuing equity-linked guaranteed minimum death benefits with European-style Asian payoffs under a regime switching jump-diffusion model

引用

COMMUNICATIONS IN NONLINEAR SCIENCE AND NUMERICAL SIMULATION 2024年 128卷

作者： Wang, Yayun Liu, Shengda Chongqing Univ Posts & Telecommun Key Lab Intelligent Anal & Decis Complex Syst Chongqing 400065 Peoples R China Chinese Acad Sci Inst Automat State Key Lab Management & Control Complex Syst Beijing 100190 Peoples R China

This paper presents a novel and efficient approach to pricing equity-linked guaranteed minimum death benefits (GMDB) with European-style geometric Asian and arithmetic Asian payoffs. Our method assumes that the underlying asset process follows a regime-switching Levy model, which captures the key features of market dynamics in the continuous transition of the economy. To derive the approximate value of GMDB products, we employ the complex Fourier series (CFS) expansion method. Our error analysis demonstrates that this approach exhibits an exponential convergence rate. In our numerical experiments, we compare the CFS approach to other Fourier transform methods and Monte Carlo simulation. The results show that our method outperforms the other approaches in terms of both efficiency and accuracy. This paper contributes to the literature on pricing equity-linked GMDB products by proposing a novel and efficient approach based on regime-switching Levy models and complex Fourier series expansion. The implications of our results may have significant practical implications for the insurance industry and financial markets.

关键词： Regime-switching Levy model complex Fourier series method European-style Asian option payoffs GMDB

来源：评论

学校读者我要写书评

暂无评论

Magnetic-optical dual-modality imaging monitoring chemotherapy efficacy of pancreatic ductal adenocarcinoma with a low-dose fibronectin-targeting Gd-based contrast agent

引用

EUROPEAN JOURNAL OF NUCLEAR MEDICINE AND MOLECULAR IMAGING 2024年第7期51卷 1841-1855页

作者： Zhang, Wenjia Liang, Xiaolong Zhang, Xinyu Tong, Wei Shi, Guangyuan Guo, Haozhuo Jin, Zhengyu Tian, Jie Du, Yang Xue, Huadan Chinese Acad Med Sci & Peking Union Med Coll Peking Union Med Coll Hosp Dept Radiol State Key Lab Complex Severe & Rare Dis 1 Shuaifuyuan Beijing 100730 Peoples R China Peking Univ Peoples Hosp Dept Radiol Beijing 100032 Peoples R China Chinese Acad Sci State Key Lab Management & Control Complex Syst Beijing Key Lab Mol Imaging CAS Key Lab Mol ImagingInst Automat 95 Zhongguancun East Rd Beijing 100190 Peoples R China Peking Univ Third Hosp Dept Ultrasound Beijing 100191 Peoples R China Beihang Univ Beijing Adv Innovat Ctr Big Data Based Precis Med Sch Med Beijing 100191 Peoples R China Univ Chinese Acad Sci Beijing 100049 Peoples R China

Purpose Pancreatic ductal adenocarcinoma (PDAC) is a lethal hypovascular tumor surrounded by dense fibrosis. Albumin-bound paclitaxel and gemcitabine (AG) chemotherapy is the mainstay of PDAC treatment through depleting peritumoral fibrosis and killing tumor cells;however, it remains challenging due to the lack of a noninvasive imaging method evaluating fibrotic changes during AG chemotherapy. In this study, we developed a dual-modality imaging platform that enables noninvasive, dynamic, and quantitative assessment of chemotherapy-induced fibrotic changes through near-infrared fluorescence molecular imaging (FMI) and magnetic resonance imaging (MRI) using an extradomain B fibronectin (EDB-FN)-targeted imaging probe (ZD2-Gd-DOTA-Cy7). Methods The ZD2-Gd-DOTA-Cy7 probe was constructed by conjugating a peptide (Cys-TVRTSAD) to Gd-DOTA and the near-infrared dye Cy7. PDAC murine xenograft models were intravenously injected with ZD2-Gd-DOTA-Cy7 at a Gd concentration of 0.05 mmol/kg or free Cy7 and Gd-DOTA as control. The normalized tumor background ratio (TBR) on FMI and the T1 reduction ratio on MRI were quantitatively analyzed. For models receiving AG chemotherapy or saline, MRI/FMI was performed before and after treatment. Histological analyses were performed for validation. Results The ZD2-Gd-DOTA-Cy7 concentration showed a linear correlation with the fluorescence intensity and T1 relaxation time in vitro. The optimal imaging time was 30 min after injection of the ZD2-Gd-DOTA-Cy7 (0.05 mmol/kg), only half of the clinic dosage of gadolinium. Additionally, ZD2-Gd-DOTA-Cy7 generated a 1.44-fold and 1.90-fold robust contrast enhancement compared with Cy7 (P < 0.05) and Gd-DOTA (P < 0.05), respectively. For AG chemotherapy monitoring, the T1 reduction ratio and normalized TBR in the fibrotic tumor areas were significantly increased by 1.99-fold (P < 0.05) and 1.78-fold (P < 0.05), respectively, in the control group compared with those in the AG group. Conclusion MRI/FMI wi

关键词： Molecular imaging Magnetic resonance imaging Fluorescence molecular imaging Pancreatic cancer Fibronectin

来源：评论

学校读者我要写书评

暂无评论

C2FDA: Coarse-to-Fine Domain Adaptation for Traffic Object Detection

引用

IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION systemS 2022年第8期23卷 12633-12647页

作者： Zhang, Hui Luo, Guiyang Li, Jinglin Wang, Fei-Yue Beijing Jiaotong Univ Sch Comp & Informat Technol Beijing 100044 Peoples R China Chinese Acad Sci Inst Automat State Key Lab Management & Control Complex Syst Beijing 100190 Peoples R China Beijing Univ Posts & Telecommun State Key Lab Networking & Switching Technol Beijing 100876 Peoples R China

Object detection in traffic scenes has attracted considerable attention from both academia and industry recently. Modern detectors achieve excellent performance under a simple constrained environment while performing poorly under the actual complex and open traffic environment. Therefore, the capability of adapting to new and unseen domains is a key factor for the large-scale application and proliferation of detectors in autonomous driving. To this end, this paper proposes a novel category-induced coarse-to-fine domain adaptation approach (C2FDA) for cross-domain object detection, which consists of three pivotal components: (1) Attention-induced coarse-grained alignment module (ACGA), which strengthens the distribution alignment across disparate domains within the foreground features in category-agnostic way by the minimax optimization between the domain classifier and the backbone feature extractor;(2) Attention-induced feature selection module, which assists the model to emphasize the crucial foreground features and enables the ACGA to focus on the relevant and discriminative foreground features, without being affected by the distribution of inconsequential background features;(3) Category-induced fine-grained alignment module (CFGA), which reduces the domain shift in category-aware way by minimizing the distance of centroids with the same category from different domains and maximizing that of centroids with disparate categories. We evaluate the performance of our approach in various source/target domain pairs and comprehensive results demonstrate that C2FDA significantly outperforms the state-of-the-art on multiple domain adaptation scenarios, i.e., the synthetic-to-real adaptation, the weather adaptation, and the cross camera adaptation.

关键词： Object detection domain adaptation synthetic data intelligent visual perception

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：