Filters








3,686 Hits in 5.2 sec

Learning Compositional Shape Priors for Few-Shot 3D Reconstruction [article]

Mateusz Michalkiewicz, Stavros Tsogkas, Sarah Parisot, Mahsa Baktashmotlagh, Anders Eriksson, Eugene Belilovsky
2021 arXiv   pre-print
In this work we experimentally demonstrate that naive baselines fail in this few-shot learning setting, in which the network must learn informative shape priors for inference of new categories.  ...  However, building large collections of 3D shapes for supervised training is a laborious process; a more realistic and less constraining task is inferring 3D shapes for categories with few available training  ...  More importantly, true 3D shape understanding im- We tackle the problem of single-view 3D reconstruction in the few-shot learning setup.  ... 
arXiv:2106.06440v2 fatcat:ncnobbj7mndbpdgmuzz6cefumq

Few-Shot Single-View 3-D Object Reconstruction with Compositional Priors [article]

Mateusz Michalkiewicz, Sarah Parisot, Stavros Tsogkas, Mahsa Baktashmotlagh, Anders Eriksson, Eugene Belilovsky
2020 arXiv   pre-print
On the other hand settings where 3D shape must be inferred for new categories with few examples are more natural and require models that generalize about shapes.  ...  In this work we demonstrate experimentally that naive baselines do not apply when the goal is to learn to reconstruct novel objects using very few examples, and that in a few-shot learning setting, the  ...  In summary, we make the following contributions: -We investigate the few-shot learning setting for 3D shape reconstruction and demonstrate that this set-up constitutes an ideal testbed for the development  ... 
arXiv:2004.06302v2 fatcat:5cphxa5awrarvn6b3brm33wzhm

Pose Adaptive Dual Mixup for Few-Shot Single-View 3D Reconstruction [article]

Ta-Ying Cheng, Hsuan-Ru Yang, Niki Trigoni, Hwann-Tzong Chen, Tyng-Luh Liu
2021 arXiv   pre-print
We present a pose adaptive few-shot learning procedure and a two-stage data interpolation regularization, termed Pose Adaptive Dual Mixup (PADMix), for single-image 3D reconstruction.  ...  PADMix significantly outperforms previous literature on few-shot settings over the ShapeNet dataset and sets new benchmarks on the more challenging real-world Pix3D dataset.  ...  Few-Shot (ICLR). Single-View 3D Reconstruction with Compositional Priors. Fan, H.; Su, H.; and Guibas, L. 2017.  ... 
arXiv:2112.12484v1 fatcat:ht6sjw4o2zex5hhfi6k4w573ne

GeLaTO: Generative Latent Textured Objects [article]

Ricardo Martin-Brualla, Rohit Pandey, Sofien Bouaziz, Matthew Brown, Dan B Goldman
2020 arXiv   pre-print
Accurate modeling of 3D objects exhibiting transparency, reflections and thin structures is an extremely challenging problem.  ...  We demonstrate the effectiveness of our approach by reconstructing complex objects from a sparse set of views.  ...  [19] learn a linear shape basis for 3D keypoints for each category, using a variant of NRSfM [6] . Kanazawa et al.  ... 
arXiv:2008.04852v1 fatcat:mfum5snn7jeh5dhas54hwyjrme

Author Index

2010 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition  
Connectivity Constraints for Reconstruction of 3D Line Segments from Images Tommasi, Tatiana Safety in Numbers: Learning Categories from Few Examples with Multi Model Knowledge Transfer Tong, Yan Workshop  ...  Georgiev, Todor Demo: Real Time Lightfield Rendering Using GPUs Gerónimo, David Learning Appearance in Virtual Scenarios for Pedestrian Detection Gevers, Theo 3D Scene Priors for Road Detection  ... 
doi:10.1109/cvpr.2010.5539913 fatcat:y6m5knstrzfyfin6jzusc42p54

Using Shape to Categorize: Low-Shot Learning with an Explicit Shape Bias [article]

Stefan Stojanov, Anh Thai, James M. Rehg
2021 arXiv   pre-print
about 3D shape can be used to improve low-shot learning methods' generalization performance.  ...  We propose a new way to improve existing low-shot learning approaches by learning a discriminative embedding space using 3D object shape, and using this embedding by learning how to map images into it.  ...  While prior work has demonstrated effective approaches to object categorization using 3D shapes as input [31, 33, 52, 53, 4] , and there is a large literature on few-shot learning from images alone [  ... 
arXiv:2101.07296v2 fatcat:qbam6yf7uvdtxaqkl6ibihnbvi

CodeNeRF: Disentangled Neural Radiance Fields for Object Categories [article]

Wonbong Jang, Lourdes Agapito
2021 arXiv   pre-print
CodeNeRF is an implicit 3D neural representation that learns the variation of object shapes and textures across a category and can be trained, from a set of posed images, to synthesize novel views of unseen  ...  Unlike the original NeRF, which is scene specific, CodeNeRF learns to disentangle shape and texture by learning separate embeddings.  ...  [34, 3, 32] to propose learned representations for 3D reconstruction with 3D supervision.  ... 
arXiv:2109.01750v1 fatcat:wllnvifnrbai3e65eqjc72uzzu

MetaSDF: Meta-learning Signed Distance Functions [article]

Vincent Sitzmann, Eric R. Chan, Richard Tucker, Noah Snavely, Gordon Wetzstein
2020 arXiv   pre-print
Generalizing across shapes with such neural implicit representations amounts to learning priors over the respective function space and enables geometry reconstruction from partial or noisy observations  ...  Here, we formalize learning of a shape space as a meta-learning problem and leverage gradient-based meta-learning algorithms to solve this task.  ...  Acknowledgments and Disclosure of Funding We would like to offer special thanks to Julien Martel, Matthew Chan, and Trevor Chan for fruitful discussions and assistance in completing this work.  ... 
arXiv:2006.09662v1 fatcat:ffasyuasirht7kwpakrfluidnm

Scene Representation Networks: Continuous 3D-Structure-Aware Neural Scene Representations [article]

Vincent Sitzmann, Michael Zollhöfer, Gordon Wetzstein
2020 arXiv   pre-print
We demonstrate the potential of SRNs by evaluating them for novel view synthesis, few-shot reconstruction, joint shape and appearance interpolation, and unsupervised discovery of a non-rigid face model  ...  This formulation naturally generalizes across scenes, learning powerful geometry and appearance priors in the process.  ...  Michael Zollhöfer was supported by the Max Planck Center for Visual Computing and Communication (MPC-VCC).  ... 
arXiv:1906.01618v2 fatcat:mdywjeydm5asbacdtbs7wmgj5u

BAE-NET: Branched Autoencoder for Shape Co-Segmentation [article]

Zhiqin Chen, Kangxue Yin, Matthew Fisher, Siddhartha Chaudhuri, Hao Zhang
2019 arXiv   pre-print
By complementing the shape reconstruction loss with a label loss, BAE-NET is easily tuned for one-shot learning.  ...  We treat shape co-segmentation as a representation learning problem and introduce BAE-NET, a branched autoencoder network, for the task.  ...  This encourages the network to learn to represent the final complex hyperdimensional shape as a composition of a few simple hyperdimensional shapes.  ... 
arXiv:1903.11228v2 fatcat:htxwsw6wovdjzmzrb2hiveb6zi

Light Field Networks: Neural Scene Representations with Single-Evaluation Rendering [article]

Vincent Sitzmann, Semon Rezchikov, William T. Freeman, Joshua B. Tenenbaum, Fredo Durand
2021 arXiv   pre-print
In the setting of simple scenes, we leverage meta-learning to learn a prior over LFNs that enables multi-view consistent light field reconstruction from as little as a single image observation.  ...  Rendering a ray from an LFN requires only a *single* network evaluation, as opposed to hundreds of evaluations per ray for ray-marching or volumetric based renderers in 3D-structured neural scene representations  ...  For the setting of simple scenes, we demonstrate that this challenge can be overcome by learning a prior over 4D light fields in a meta-learning framework.  ... 
arXiv:2106.02634v1 fatcat:niopcbm5cjhw5iwih4uv27oo54

Self-supervised Tumor Segmentation through Layer Decomposition [article]

Xiaoman Zhang, Weidi Xie, Chaoqin Huang, Yanfeng Wang, Ya Zhang, Xin Chen, Qi Tian
2021 arXiv   pre-print
In this paper, we target self-supervised representation learning for zero-shot tumor segmentation.  ...  Forth, our approach achieves superior results for zero-shot tumor segmentation on different downstream datasets, BraTS2018 for brain tumor segmentation and LiTS2017 for liver tumor segmentation.  ...  Network Architecture Our proposed self-supervised learning method is flexible for any 3D encoder-decoder network.  ... 
arXiv:2109.03230v3 fatcat:xaxxtxopzjgxzhvty4psfnswjy

Disentangling 3D Prototypical Networks For Few-Shot Concept Learning [article]

Mihir Prabhudesai, Shamit Lal, Darshan Patil, Hsiao-Yu Tung, Adam W Harley, Katerina Fragkiadaki
2021 arXiv   pre-print
We present neural architectures that disentangle RGB-D images into objects' shapes and styles and a map of the background scene, and explore their applications for few-shot 3D object detection and few-shot  ...  We show that the proposed 3D neural representations are compositional: they can generate novel 3D scene feature maps by mixing object shapes and styles, resizing and adding the resulting object 3D feature  ...  RELATION TO PREVIOUS WORKS Few-shot concept learning Few-shot learning methods attempt to learn a new concept from one or a few annotated examples at test time, yet, at training time, these models still  ... 
arXiv:2011.03367v3 fatcat:bqxpkyamcnf53cvbgrravlgk4m

Flow Guided Transformable Bottleneck Networks for Motion Retargeting [article]

Jian Ren, Menglei Chai, Oliver J. Woodford, Kyle Olszewski, Sergey Tulyakov
2021 arXiv   pre-print
Few-shot motion transfer techniques, which only require one or a few images from a target, have recently drawn considerable attention.  ...  This allows us to learn our 3D representation solely from videos of moving people.  ...  Conclusion Our approach to few-shot human motion retargeting exploits advantages of 3D representations of human body while avoiding limitations of the more straightforward prior methods.  ... 
arXiv:2106.07771v1 fatcat:yjprgqyphbdenkfxguq5dpysxu

Im2Struct: Recovering 3D Shape Structure from a Single RGB Image [article]

Chengjie Niu, Jun Li, Kai Xu
2018 arXiv   pre-print
We propose to recover 3D shape structures from single RGB images, where structure refers to shape parts represented by cuboids and part relations encompassing connectivity and symmetry.  ...  We demonstrate two applications of our method including structure-guided completion of 3D volumes reconstructed from single-view images and structure-aware interactive editing of 2D images.  ...  Human brains do well both in shape inference based on low-level visual stimulus and structural reasoning with the help of prior knowledge about 3D shape compositions.  ... 
arXiv:1804.05469v1 fatcat:2tqr7m5o3fhwlpbdrgso7gwcia
« Previous Showing results 1 — 15 out of 3,686 results