Computer Graphics Forum最新文献

筛选
英文 中文
Front Matter 前言
IF 2.7 4区 计算机科学
Computer Graphics Forum Pub Date : 2024-11-18 DOI: 10.1111/cgf.15188
{"title":"Front Matter","authors":"","doi":"10.1111/cgf.15188","DOIUrl":"https://doi.org/10.1111/cgf.15188","url":null,"abstract":"<p>McGill University, Montreal, Canada</p><p>August 21 to August 23, 2024</p><p><b>Conference Co-Chairs</b></p><p>Paul Kry, McGill University</p><p>Marie-Paule Cani, École Polytechnique</p><p><b>Program Co-Chairs</b></p><p>Melina Skouras, Inria Grenoble Rhone-Alpes</p><p>He Wang, University College London</p><p><b>Poster Chair</b></p><p>Victor Zordan, Roblox</p><p>\u0000 </p><p>Ando, Ryoichi – Unaffiliated</p><p>Andrews, Sheldon – École de technologie supérieure</p><p>Aristidou, Andreas – University of Cyprus & CYENS Centre of Excellence</p><p>Barbic, Jernej – University of Southern California</p><p>Batty, Christopher – University of Waterloo</p><p>Bender, Jan – RWTH Aachen University</p><p>Benes, Bedrich – Purdue University</p><p>Bickel, Bernd – IST Austria</p><p>Chen, Peter Yichen – Massachusetts Institute of Technology</p><p>Chen, Zhen – UT Austin</p><p>Chentanez, Nuttapong – NVIDIA</p><p>Chu, Mengyu – Peking University</p><p>Cirio, Gabriel – SEDDI</p><p>Deng, Zhigang – University of Houston</p><p>Durupinar Babur, Funda – UMASS Boston</p><p>Erleben, Kenny – Department of Computer Science, University of Copenhagen</p><p>Grinspun, Eitan – University of Toronto</p><p>He, Feixiang – University College London</p><p>Ho, Edmond S. L. – University of Glasgow</p><p>Holden, Daniel – Epic Games</p><p>Hoyet, Ludovic – Centre Inria de l'Université de Rennes</p><p>Jiang, Chenfanfu – UCLA</p><p>Jin, Xiaogang – State Key Lab of CAD&CG, Zhejiang University</p><p>Kapadia, Mubbasir – Rutgers</p><p>Kaufman, Danny – Adobe Research</p><p>Kim, Theodore – Yale University</p><p>Langlois, Timothy – Adobe</p><p>Lee, Sung-Hee – KAIST</p><p>Li, Jing – University of Utah</p><p>Li, Minchen – Carnegie Mellon University</p><p>Liu, Libin – Peking University</p><p>Liu, Tiantian – Taichi Graphics</p><p>Ly, Mickaël – IST Austria</p><p>Marchal, Maud – IRISA/INSA</p><p>Michels, Dominik – KAUST</p><p>Michiel, van de Panne – University of British Columbia</p><p>Musse, Soraia – PUCRS</p><p>Narain, Rahul – Indian Institute of Technology Delhi</p><p>Neff, Michael – University of California, Davis</p><p>Otaduy, Miguel A. – Universidad Rey Juan Carlos, Madrid</p><p>Pai, Dinesh – University of British Columbia</p><p>Pelechano, Nuria – Universitat Politecnica de Catalunya</p><p>Pettre, Julien – Inria</p><p>Pollard, Nancy – Carnegie Mellon University</p><p>Popa, Tiberiu – Concordia University</p><p>Rohmer, Damien – Ecole Polytechnique</p><p>Schreck, Camille – Inria Nancy</p><p>Shinar, Tamar – UC RIVERSIDE</p><p>Stomakhin, Alexey – Weta Digital</p><p>Sueda, Shinjiro – Texas A&M University</p><p>Tang, Min – Zhejiang University</p><p>Teschner, Matthias – University of Freiburg</p><p>Thomaszewski, Bernhard – ETH Zurich</p><p>Umetani, Nobuyuki – The University of Tokyo</p><p>Vouga, Etienne – UT Austin</p><p>Wang, Stephanie – Independent Researcher</p><p>Wang, Yingying – McMaster University</p><p>Wojtan, Chris – Institute of Science and Technology Austria (ISTA)</p><p>Won, Jungda","PeriodicalId":10687,"journal":{"name":"Computer Graphics Forum","volume":"43 8","pages":"i-ix"},"PeriodicalIF":2.7,"publicationDate":"2024-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1111/cgf.15188","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142707771","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
Front Matter 前言
IF 2.7 4区 计算机科学
Computer Graphics Forum Pub Date : 2024-11-14 DOI: 10.1111/cgf.14853
{"title":"Front Matter","authors":"","doi":"10.1111/cgf.14853","DOIUrl":"https://doi.org/10.1111/cgf.14853","url":null,"abstract":"<p>The 32nd Pacific Conference on Computer Graphics and Applications</p><p>Huangshan (Yellow Mountain), China</p><p>October 13 – 16, 2024</p><p><b>Conference Co-Chairs</b></p><p>Jan Bender, RWTH Aachen, Germany</p><p>Ligang Liu, University of Science and Technology of China, China</p><p>Denis Zorin, New York University, USA</p><p><b>Program Co-Chairs</b></p><p>Renjie Chen, University of Science and Technology of China, China</p><p>Tobias Ritschel, University College London, UK</p><p>Emily Whiting, Boston University, USA</p><p><b>Organization Co-Chairs</b></p><p>Xiao-Ming Fu, University of Science and Technology of China, China</p><p>Jianwei Hu, Huangshan University, China</p><p>The 2024 Pacific Graphics Conference, held in the scenic city of Huangshan, China from October 13-16, marked a milestone year with record-breaking participation and submissions. As one of the premier forums for computer graphics research, the conference maintained its high standards of academic excellence while taking measures to handle unprecedented submission volumes.</p><p>This year saw an extraordinary 360 full paper submissions, the highest in Pacific Graphics history. To maintain our rigorous review standards, we implemented a streamlined process including an initial sorting committee and desk review phase. Of the 305 submissions that proceeded to full review, each received a minimum of 3 reviews, with an average of 3.76 reviews per submission. Our double-blind review process was managed by an International Program Committee (IPC) comprising 112 experts, carefully selected to ensure regular renewal of perspectives in the field.</p><p>In the review process, each submission was assigned to two IPC members as primary and secondary reviewers. These reviewers, in turn, invited two additional tertiary reviewers, ensuring comprehensive evaluation. Authors were provided a five-day window to submit 1,000-word rebuttals addressing reviewer comments and potential misunderstandings. This year's IPC meeting was conducted virtually over one week through asynchronous discussions.</p><p>From the initial 360 submissions, 109 papers were conditionally accepted, yielding an acceptance rate of 30.28%. Following the acceptance notifications, resulting in a final publication count of 105 papers. These were distributed across publication venues as follows: 59 papers were selected for journal publication in Computer Graphics Forum, while 50 papers were accepted to the Conference Track and published in the Proceedings. Additionally, 6 papers were recommended for fast-track review with major revisions for future Computer Graphics Forum consideration.</p><p>The accepted papers showcase the breadth of modern computer graphics research, spanning computational photography, geometry and mesh processing, appearance, shading, texture, rendering technologies, 3D scanning and analysis, physical simulation, human animation and motion capture, crowd and cloth simulation, 3D printing and fabrication, dig","PeriodicalId":10687,"journal":{"name":"Computer Graphics Forum","volume":"43 7","pages":"i-xxii"},"PeriodicalIF":2.7,"publicationDate":"2024-11-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1111/cgf.14853","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142664863","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
DiffPop: Plausibility-Guided Object Placement Diffusion for Image Composition DiffPop:用于图像合成的似是而非引导的物体位置扩散
IF 2.7 4区 计算机科学
Computer Graphics Forum Pub Date : 2024-11-14 DOI: 10.1111/cgf.15246
Jiacheng Liu, Hang Zhou, Shida Wei, Rui Ma
{"title":"DiffPop: Plausibility-Guided Object Placement Diffusion for Image Composition","authors":"Jiacheng Liu,&nbsp;Hang Zhou,&nbsp;Shida Wei,&nbsp;Rui Ma","doi":"10.1111/cgf.15246","DOIUrl":"https://doi.org/10.1111/cgf.15246","url":null,"abstract":"<p>In this paper, we address the problem of plausible object placement for the challenging task of realistic image composition. We propose DiffPop, the first framework that utilizes plausibility-guided denoising diffusion probabilistic model to learn the scale and spatial relations among multiple objects and the corresponding scene image. First, we train an unguided diffusion model to directly learn the object placement parameters in a self-supervised manner. Then, we develop a human-in-the-loop pipeline which exploits human labeling on the diffusion-generated composite images to provide the weak supervision for training a structural plausibility classifier. The classifier is further used to guide the diffusion sampling process towards generating the plausible object placement. Experimental results verify the superiority of our method for producing plausible and diverse composite images on the new Cityscapes-OP dataset and the public OPA dataset, as well as demonstrate its potential in applications such as data augmentation and multi-object placement tasks. Our dataset and code will be released.</p>","PeriodicalId":10687,"journal":{"name":"Computer Graphics Forum","volume":"43 7","pages":""},"PeriodicalIF":2.7,"publicationDate":"2024-11-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142664803","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
iShapEditing: Intelligent Shape Editing with Diffusion Models iShapEditing:利用扩散模型进行智能形状编辑
IF 2.7 4区 计算机科学
Computer Graphics Forum Pub Date : 2024-11-08 DOI: 10.1111/cgf.15253
Jing Li, Juyong Zhang, Falai Chen
{"title":"iShapEditing: Intelligent Shape Editing with Diffusion Models","authors":"Jing Li,&nbsp;Juyong Zhang,&nbsp;Falai Chen","doi":"10.1111/cgf.15253","DOIUrl":"https://doi.org/10.1111/cgf.15253","url":null,"abstract":"<p>Recent advancements in generative models have enabled image editing very effective with impressive results. By extending this progress to 3D geometry models, we introduce iShapEditing, a novel framework for 3D shape editing which is applicable to both generated and real shapes. Users manipulate shapes by dragging handle points to corresponding targets, offering an intuitive and intelligent editing interface. Leveraging the Triplane Diffusion model and robust intermediate feature correspondence, our framework utilizes classifier guidance to adjust noise representations during sampling process, ensuring alignment with user expectations while preserving plausibility. For real shapes, we employ shape predictions at each time step alongside a DDPM-based inversion algorithm to derive their latent codes, facilitating seamless editing. iShapEditing provides effective and intelligent control over shapes without the need for additional model training or fine-tuning. Experimental examples demonstrate the effectiveness and superiority of our method in terms of editing accuracy and plausibility.</p>","PeriodicalId":10687,"journal":{"name":"Computer Graphics Forum","volume":"43 7","pages":""},"PeriodicalIF":2.7,"publicationDate":"2024-11-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142664546","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
𝒢-Style: Stylized Gaussian Splatting 𝒢-风格:风格化高斯拼接
IF 2.7 4区 计算机科学
Computer Graphics Forum Pub Date : 2024-11-08 DOI: 10.1111/cgf.15259
Áron Samuel Kovács, Pedro Hermosilla, Renata G. Raidou
{"title":"𝒢-Style: Stylized Gaussian Splatting","authors":"Áron Samuel Kovács,&nbsp;Pedro Hermosilla,&nbsp;Renata G. Raidou","doi":"10.1111/cgf.15259","DOIUrl":"https://doi.org/10.1111/cgf.15259","url":null,"abstract":"<div>\u0000 \u0000 <p>We introduce 𝒢-Style, a novel algorithm designed to transfer the style of an image onto a 3D scene represented using Gaussian Splatting. Gaussian Splatting is a powerful 3D representation for novel view synthesis, as—compared to other approaches based on Neural Radiance Fields—it provides fast scene renderings and user control over the scene. Recent pre-prints have demonstrated that the style of Gaussian Splatting scenes can be modified using an image exemplar. However, since the scene geometry remains fixed during the stylization process, current solutions fall short of producing satisfactory results. Our algorithm aims to address these limitations by following a three-step process: In a pre-processing step, we remove undesirable Gaussians with large projection areas or highly elongated shapes. Subsequently, we combine several losses carefully designed to preserve different scales of the style in the image, while maintaining as much as possible the integrity of the original scene content. During the stylization process and following the original design of Gaussian Splatting, we split Gaussians where additional detail is necessary within our scene by tracking the gradient of the stylized color. Our experiments demonstrate that 𝒢-Style generates high-quality stylizations within just a few minutes, outperforming existing methods both qualitatively and quantitatively.</p>\u0000 </div>","PeriodicalId":10687,"journal":{"name":"Computer Graphics Forum","volume":"43 7","pages":""},"PeriodicalIF":2.7,"publicationDate":"2024-11-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1111/cgf.15259","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142664544","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
LGSur-Net: A Local Gaussian Surface Representation Network for Upsampling Highly Sparse Point Cloud LGSur-Net:用于高稀疏点云升采样的局部高斯曲面表示网络
IF 2.7 4区 计算机科学
Computer Graphics Forum Pub Date : 2024-11-08 DOI: 10.1111/cgf.15257
Zijian Xiao, Tianchen Zhou, Li Yao
{"title":"LGSur-Net: A Local Gaussian Surface Representation Network for Upsampling Highly Sparse Point Cloud","authors":"Zijian Xiao,&nbsp;Tianchen Zhou,&nbsp;Li Yao","doi":"10.1111/cgf.15257","DOIUrl":"https://doi.org/10.1111/cgf.15257","url":null,"abstract":"<p>We introduce LGSur-Net, an end-to-end deep learning architecture, engineered for the upsampling of sparse point clouds. LGSur-Net harnesses a trainable Gaussian local representation by positioning a series of Gaussian functions on an oriented plane, complemented by the optimization of individual covariance matrices. The integration of parametric factors allows for the encoding of the plane's rotational dynamics and Gaussian weightings into a linear transformation matrix. Then we extract the feature maps from the point cloud and its adjoining edges and learn the local Gaussian depictions to accurately model the shape's local geometry through an attention-based network. The Gaussian representation's inherent high-order continuity endows LGSur-Net with the natural ability to predict surface normals and support upsampling to any specified resolution. Comprehensive experiments validate that LGSur-Net efficiently learns from sparse data inputs, surpassing the performance of existing state-of-the-art upsampling methods. Our code is publicly available at https://github.com/Rangiant5b72/LGSur-Net.</p>","PeriodicalId":10687,"journal":{"name":"Computer Graphics Forum","volume":"43 7","pages":""},"PeriodicalIF":2.7,"publicationDate":"2024-11-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142664543","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
Cinematic Gaussians: Real-Time HDR Radiance Fields with Depth of Field 电影般的高斯具有景深的实时 HDR 辐射场
IF 2.7 4区 计算机科学
Computer Graphics Forum Pub Date : 2024-11-07 DOI: 10.1111/cgf.15214
Chao Wang, Krzysztof Wolski, Bernhard Kerbl, Ana Serrano, Mojtaba Bemana, Hans-Peter Seidel, Karol Myszkowski, Thomas Leimkühler
{"title":"Cinematic Gaussians: Real-Time HDR Radiance Fields with Depth of Field","authors":"Chao Wang,&nbsp;Krzysztof Wolski,&nbsp;Bernhard Kerbl,&nbsp;Ana Serrano,&nbsp;Mojtaba Bemana,&nbsp;Hans-Peter Seidel,&nbsp;Karol Myszkowski,&nbsp;Thomas Leimkühler","doi":"10.1111/cgf.15214","DOIUrl":"https://doi.org/10.1111/cgf.15214","url":null,"abstract":"<div>\u0000 <p>Radiance field methods represent the state of the art in reconstructing complex scenes from multi-view photos. However, these reconstructions often suffer from one or both of the following limitations: First, they typically represent scenes in low dynamic range (LDR), which restricts their use to evenly lit environments and hinders immersive viewing experiences. Secondly, their reliance on a pinhole camera model, assuming all scene elements are in focus in the input images, presents practical challenges and complicates refocusing during novel-view synthesis. Addressing these limitations, we present a lightweight method based on 3D Gaussian Splatting that utilizes multi-view LDR images of a scene with varying exposure times, apertures, and focus distances as input to reconstruct a high-dynamic-range (HDR) radiance field. By incorporating analytical convolutions of Gaussians based on a thin-lens camera model as well as a tonemapping module, our reconstructions enable the rendering of HDR content with flexible refocusing capabilities. We demonstrate that our combined treatment of HDR and depth of field facilitates real-time cinematic rendering, outperforming the state of the art.</p>\u0000 </div>","PeriodicalId":10687,"journal":{"name":"Computer Graphics Forum","volume":"43 7","pages":""},"PeriodicalIF":2.7,"publicationDate":"2024-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1111/cgf.15214","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142664606","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
GETr: A Geometric Equivariant Transformer for Point Cloud Registration GETr:用于点云注册的几何等差变换器
IF 2.7 4区 计算机科学
Computer Graphics Forum Pub Date : 2024-11-07 DOI: 10.1111/cgf.15216
Chang Yu, Sanguo Zhang, Li-Yong Shen
{"title":"GETr: A Geometric Equivariant Transformer for Point Cloud Registration","authors":"Chang Yu,&nbsp;Sanguo Zhang,&nbsp;Li-Yong Shen","doi":"10.1111/cgf.15216","DOIUrl":"https://doi.org/10.1111/cgf.15216","url":null,"abstract":"<div>\u0000 \u0000 <p>As a fundamental problem in computer vision, 3D point cloud registration (PCR) aims to seek the optimal transformation to align point cloud pairs. Meanwhile, the equivariance lies at the core of matching point clouds at arbitrary pose. In this paper, we propose GETr, a geometric equivariant transformer for PCR. By learning the point-wise orientations, we decouple the coordinate to the pose of the point clouds, which is the key to achieve equivariance in our framework. Then we utilize attention mechanism to learn the geometric features for superpoints matching, the proposed novel self-attention mechanism encodes the geometric information of point clouds. Finally, the coarse-to-fine manner is used to obtain high-quality correspondence for registration. Extensive experiments on both indoor and outdoor benchmarks demonstrate that our method outperforms various existing state-of-the-art methods.</p>\u0000 </div>","PeriodicalId":10687,"journal":{"name":"Computer Graphics Forum","volume":"43 7","pages":""},"PeriodicalIF":2.7,"publicationDate":"2024-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1111/cgf.15216","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142664645","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
Strictly Conservative Neural Implicits 严格保守的神经暗示
IF 2.7 4区 计算机科学
Computer Graphics Forum Pub Date : 2024-11-07 DOI: 10.1111/cgf.15241
I. Ludwig, M. Campen
{"title":"Strictly Conservative Neural Implicits","authors":"I. Ludwig,&nbsp;M. Campen","doi":"10.1111/cgf.15241","DOIUrl":"https://doi.org/10.1111/cgf.15241","url":null,"abstract":"<div>\u0000 \u0000 <p>We describe a method to convert 3D shapes into neural implicit form such that the shape is approximated in a guaranteed conservative manner. This means the input shape is strictly contained inside the neural implicit or, alternatively, vice versa. Such conservative approximations are of interest in a variety of applications, including collision detection, occlusion culling, or intersection testing. Our approach is the first to guarantee conservativeness in this context of neural implicits. We support input given as mesh, voxel set, or implicit function. Adaptive affine arithmetic is employed in the neural network fitting process, enabling the reasoning over infinite sets of points despite using a finite set of training data. Combined with an interior point style optimization approach this yields the desired guarantee.</p>\u0000 </div>","PeriodicalId":10687,"journal":{"name":"Computer Graphics Forum","volume":"43 7","pages":""},"PeriodicalIF":2.7,"publicationDate":"2024-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://onlinelibrary.wiley.com/doi/epdf/10.1111/cgf.15241","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142664670","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
SCARF: Scalable Continual Learning Framework for Memory-efficient Multiple Neural Radiance Fields SCARF:记忆高效多重神经辐射场的可扩展连续学习框架
IF 2.7 4区 计算机科学
Computer Graphics Forum Pub Date : 2024-11-07 DOI: 10.1111/cgf.15255
Yuze Wang, Junyi Wang, Chen Wang, Wantong Duan, Yongtang Bao, Yue Qi
{"title":"SCARF: Scalable Continual Learning Framework for Memory-efficient Multiple Neural Radiance Fields","authors":"Yuze Wang,&nbsp;Junyi Wang,&nbsp;Chen Wang,&nbsp;Wantong Duan,&nbsp;Yongtang Bao,&nbsp;Yue Qi","doi":"10.1111/cgf.15255","DOIUrl":"https://doi.org/10.1111/cgf.15255","url":null,"abstract":"<p>This paper introduces a novel continual learning framework for synthesising novel views of multiple scenes, learning multiple 3D scenes incrementally, and updating the network parameters only with the training data of the upcoming new scene. We build on Neural Radiance Fields (NeRF), which uses multi-layer perceptron to model the density and radiance field of a scene as the implicit function. While NeRF and its extensions have shown a powerful capability of rendering photo-realistic novel views in a single 3D scene, managing these growing 3D NeRF assets efficiently is a new scientific problem. Very few works focus on the efficient representation or continuous learning capability of multiple scenes, which is crucial for the practical applications of NeRF. To achieve these goals, our key idea is to represent multiple scenes as the linear combination of a cross-scene weight matrix and a set of scene-specific weight matrices generated from a global parameter generator. Furthermore, we propose an uncertain surface knowledge distillation strategy to transfer the radiance field knowledge of previous scenes to the new model. Representing multiple 3D scenes with such weight matrices significantly reduces memory requirements. At the same time, the uncertain surface distillation strategy greatly overcomes the catastrophic forgetting problem and maintains the photo-realistic rendering quality of previous scenes. Experiments show that the proposed approach achieves state-of-the-art rendering quality of continual learning NeRF on NeRF-Synthetic, LLFF, and TanksAndTemples datasets while preserving extra low storage cost.</p>","PeriodicalId":10687,"journal":{"name":"Computer Graphics Forum","volume":"43 7","pages":""},"PeriodicalIF":2.7,"publicationDate":"2024-11-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142664672","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
引用次数: 0
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
相关产品
×
本文献相关产品
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信