检索结果-南通市图书馆

Neural 3D reconstruction from sparse views using geometric priors

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Computational Visual Media 2023年第4期9卷 687-697页

作者： tai-jiang mu Hao-Xiang Chen Jun-Xiong Cai Ning Guo BNRist Department of Computer Science and TechnologyTsinghua UniversityBeijing 100084China Academy of Military Sciences Beijing 100091China

Sparse view 3D reconstruction has attracted increasing attention with the development of neural implicit 3D *** methods usually only make use of 2D views,requiring a dense set of input views for accurate 3D *** this paper,we show that accurate 3D reconstruction can be achieved by incorporating geometric priors into neural implicit 3D *** method adopts the signed distance function as the 3D representation,and learns a generalizable 3D surface reconstruction model from sparse ***,we build a more effective and sparse feature volume from the input views by using corresponding depth maps,which can be provided by depth sensors or directly predicted from the input *** recover better geometric details by imposing both depth and surface normal constraints in addition to the color loss when training the neural implicit 3D *** demonstrate that our method both outperforms state-of-the-art approaches,and achieves good generalizability.

关键词： sparse views 3D reconstruction volume rendering geometric priors neural implicit 3D representation

Attention mechanisms in computer vision:A survey

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Computational Visual Media 2022年第3期8卷 331-368页

作者： Meng-Hao Guo Tian-Xing Xu jiang-jiang Liu Zheng-Ning Liu Peng-Tao jiang tai-jiang mu Song-Hai Zhang Ralph R.Martin Ming-Ming Cheng Shi-Min Hu BNRist Department of Computer Science and TechnologyTsinghua UniversityBeijing 100084China TKLNDST College of Computer ScienceNankai UniversityTianjin 300350China School of Computer Science and Informatics Cardiff UniversityCardiffUK

Humans can naturally and effectively find salient regions in complex *** by this observation,attention mechanisms were introduced into computer vision with the aim of imitating this aspect of the human visual *** an attention mechanism can be regarded as a dynamic weight adjustment process based on features of the input *** mechanisms have achieved great success in many visual tasks,including image classification,object detection,semantic segmentation,video understanding,image generation,3D vision,multimodal tasks,and self-supervised *** this survey,we provide a comprehensive review of various attention mechanisms in computer vision and categorize them according to approach,such as channel attention,spatial attention,temporal attention,and branch attention;a related repository https://***/MenghaoG uo/Awesome-Vision-Attentions is dedicated to collecting related *** also suggest future directions for attention mechanism research.

关键词： attention transformer computer vision deep learning salience

Detecting human-object interaction with multi-level pairwise feature network

在线全文

学校读者我要写书评

暂无评论

Computational Visual Media 2021年第2期7卷 229-239页

作者： Hanchao Liu tai-jiang mu Xiaolei Huan Key Laboratory of Pervasive Computing Ministry of EducationBNRistDepartment of Computer Science and TechnologyTsinghua UniversityBeijing 100084China College of Information Sciences and Technology Pennsylvania State UniversityUniversity ParkPA 16802USA.

Human–object interaction(HOI)detection is crucial for human-centric image understanding which aims to infer human,action,object triplets within an *** studies often exploit visual features and the spatial configuration of a human–object pair in order to learn the action linking the human and object in the *** argue that such a paradigm of pairwise feature extraction and action inference can be applied not only at the whole human and object instance level,but also at the part level at which a body part interacts with an object,and at the semantic level by considering the semantic label of an object along with human appearance and human–object spatial configuration,to infer the *** thus propose a multi-level pairwise feature network(PFNet)for detecting human–object *** network consists of three parallel streams to characterize HOI utilizing pairwise features at the above three levels;the three streams are finally fused to give the action *** experiments show that our proposed PFNet outperforms other state-of-the-art methods on the VCOCO dataset and achieves comparable results to the state-of-the-art on the HICO-DET dataset.

关键词： human–object interaction detection pairwise feature network deep learning multi-level object instance

Synthesizing Robot Programs with Interactive Tutor Mode

在线全文

学校读者我要写书评

暂无评论

International Journal of Automation and computing 2019年第4期16卷 462-474页

作者： Hao Li Yu-Ping Wang tai-jiang mu Tsinghua National Laboratory for Information Science and Technology(TNList) Department of Computer Science and TechnologyTsinghua University

With the rapid development of the robotic industry, domestic robots have become increasingly popular. As domestic robots are expected to be personal assistants, it is important to develop a natural language-based human-robot interactive system for end-users who do not necessarily have much programming knowledge. To build such a system, we developed an interactive tutoring framework, named " Holert”, which can translate task descriptions in natural language to machine-interpretable logical forms automatically. Compared to previous works, Holert allows users to teach the robot by further explaining their intentions in an interactive tutor mode. Furthermore, Holert introduces a semantic dependency model to enable the robot to " understand” similar task descriptions. We have deployed Holert on an open-source robot platform, Turtlebot 2. Experimental results show that the system accuracy could be significantly improved by 163.9% with the support of the tutor mode. This system is also efficient. Even the longest task session with 10 sentences can be handled within 0.7 s.

关键词： Human-robot interaction semantic parsing program synthesis intelligent robotic systems natural language unders-tanding

HDR-Net-Fusion:Real-time 3D dynamic scene reconstruction with a hierarchical deep reinforcement network

在线全文

学校读者我要写书评

暂无评论

Computational Visual Media 2021年第4期7卷 419-435页

作者： Hao-Xuan Song Jiahui Huang Yan-Pei Cao tai-jiang mu BNRist Department of Computer Science and TechnologyTsinghua UniversityBeiing 100084China Kuaishou Technology Co. Ltd.Beijing 100085China

Reconstructing dynamic scenes with commodity depth cameras has many applications in computer graphics,computer vision,and ***,due to the presence of noise and erroneous observations from data capturing devices and the inherently ill-posed nature of non-rigid registration with insufficient information,traditional approaches often produce low-quality geometry with holes,bumps,and *** propose a novel 3D dynamic reconstruction system,named HDR-Net-Fusion,which learns to simultaneously reconstruct and refine the geometry on the fly with a sparse embedded deformation graph of surfels,using a hierarchical deep reinforcement(HDR)*** latter comprises two parts:a global HDR-Net which rapidly detects local regions with large geometric errors,and a local HDR-Net serving as a local patch refinement operator to promptly complete and enhance such *** the global HDR-Net is formulated as a novel reinforcement learning problem to implicitly learn the region selection strategy with the goal of improving the overall reconstruction *** applicability and efficiency of our approach are demonstrated using a large-scale dynamic reconstruction *** method can reconstruct geometry with higher quality than traditional methods.

关键词： dynamic 3D scene reconstruction deep reinforcement learning point cloud completion deep neural networks

A new dataset of dog breed images and a benchmark for fine-grained classification

在线全文

学校读者我要写书评

暂无评论

Computational Visual Media 2020年第4期6卷 477-487页

作者： Ding-Nan Zou Song-Hai Zhang tai-jiang mu Min Zhang Department of Computer Science and Technology BNRistTsinghua UniversityBeijing 100084China NaJiu Company Hunan 410022China Harvard Medical School Brigham and Women’s HospitalBostonMA 02115USA

In this paper, we introduce an image dataset for fine-grained classification of dog breeds: the Tsinghua Dogs Dataset. It is currently the largest dataset for fine-grained classification of dogs, including 130 dog breeds and 70,428 real-world images. It has only one dog in each image and provides annotated bounding boxes for the whole body and head. In comparison to previous similar datasets, it contains more breeds and more carefully chosen images for each breed. The diversity within each breed is greater,with between 200 and 7000+ images for each *** of the whole body and head makes the dataset not only suitable for the improvement of finegrained image classification models based on overall features, but also for those locating local informative parts. We show that dataset provides a tough challenge by benchmarking several state-of-the-art deep neural models. The dataset is available for academic purposes at https://***/ThuDogs/.

关键词： fine-grained classification dog dataset benchmark

SpinNet: Spinning convolutional network for lane boundary detection

在线全文

学校读者我要写书评

暂无评论

Computational Visual Media 2019年第4期5卷 417-428页

作者： Ruochen Fan Xuanrun Wang Qibin Hou Hanchao Liu tai-jiang mu Tsinghua University Beijing100084China Nankai University Tianjin300350China

In this paper,we propose a simple but effective framework for lane boundary detection,called Spin *** that cars or pedestrians often occlude lane boundaries and that the local features of lane boundaries are not distinctive,therefore,analyzing and collecting global context information is crucial for lane boundary *** this end,we design a novel spinning convolution layer and a brand-new lane parameterization branch in our network to detect lane boundaries from a global *** extract features in narrow strip-shaped fields,we adopt stripshaped convolutions with kernels which have 1×n or n×1 shape in the spinning convolution *** tackle the problem of that straight strip-shaped convolutions are only able to extract features in vertical or horizontal directions,we introduce the concept of feature map rotation to allow the convolutions to be applied in multiple directions so that more information can be collected concerning a whole lane ***,unlike most existing lane boundary detectors,which extract lane boundaries from segmentation masks,our lane boundary parameterization branch predicts a curve expression for the lane boundary for each pixel in the output feature *** the network utilizes this information to predict the weights of the curve,to better form the final lane *** framework is easy to implement and end-to-end *** show that our proposed Spin Net outperforms state-of-the-art methods.

关键词： object detection lane boundary detection autonomous driving deep learning

A Large Chinese Text Dataset in the Wild

在线全文

学校读者我要写书评

暂无评论

Journal of Computer Science & Technology 2019年第3期34卷 509-521页

作者： tai-Ling Yuan Zhe Zhu Kun Xu Cheng-Jun Li tai-jiang mu Shi-Min Hu Department of Computer Science and Technology Tsinghua UniversityBeijing 100084China Department of Radiology Duke UniversityNorth Carolina 27708U.S.A. Tencent Technology(Beijing)Co.Ltd. Beijing 100080China

In this paper,we introduce a very large Chinese text dataset,in the *** optical character recognition(OCR)in document images is well studied and many commercial tools are available,the detection and recognition of text in natural images is still a challenging problem,especially for some more complicated character sets such as Chinese *** of training data has always been a problem,especially for deep learning methods which require massive training *** this paper,we provide details of a newly created dataset of Chinese text with about 1 million Chinese characters from 3850 unique ones annotated by experts in over 30 000 street view *** is a challenging dataset with good diversity containing planar text,raised text,text under poor illumination,distant text,partially occluded text,*** each character,the annotation includes its underlying character,bounding box,and six *** attributes indicate the charactcr's background complexity,appearance,style,*** the dataset,we give baseline results using state-of-the-art methods for tliree tasks:character recognition(top-1 accuracy of 80.5%),character detection(AP of 70.9%),and text line detection(AED of 22.1).The dataset,source code,and trained models are publicly available.

关键词： Chinese text dataset Chinese text detection Chinese text recognition

Lane Detection:A Survey with New Results

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Journal of Computer Science & Technology 2020年第3期35卷 493-505页

作者： Dun Liang Yuan-Chen Guo Shao-Kui Zhang tai-jiang mu Xiaolei Huang Department of Computer Science and Technology BNRistTsinghua UniversityBeijing100084China College of Information Sciences and Technology Pennsylvania State UniversityUniversity ParkPA16802USA

Lane detection is essential for many aspects of autonomous driving,such as lane-based navigation and high-definition(HD)map *** lane detection is challenging especially with complex road conditions,considerable progress has been witnessed in this area in the past several *** this survey,we review recent visual-based lane detection datasets and *** datasets,we categorize them by annotations,provide detailed descriptions for each category,and show comparisons among *** methods,we focus on methods based on deep learning and organize them in terms of their detection ***,we introduce a new dataset with more detailed annotations for HD map modeling,a new direction for lane detection that is applicable to autonomous driving in complex road conditions,a deep neural network LineNet for lane detection,and show its application to HD map modeling.

关键词： convolutional neural network(CNN) dataset deep learning high-definition(HD)map lane detection