Matches in SemOpenAlex for { <https://semopenalex.org/work/W4366552803> ?p ?o ?g. }
Showing items 1 to 77 of
77
with 100 items per page.
- W4366552803 abstract "Visual object recognition systems need to generalize from a set of 2D training views to novel views. The question of how the human visual system can generalize to novel views has been studied and modeled in psychology, computer vision, and neuroscience. Modern deep learning architectures for object recognition generalize well to novel views, but the mechanisms are not well understood. In this paper, we characterize the ability of common deep learning architectures to generalize to novel views. We formulate this as a supervised classification task where labels correspond to unique 3D objects and examples correspond to 2D views of the objects at different 3D orientations. We consider three common models of generalization to novel views: (i) full 3D generalization, (ii) pure 2D matching, and (iii) matching based on a linear combination of views. We find that deep models generalize well to novel views, but they do so in a way that differs from all these existing models. Extrapolation to views beyond the range covered by views in the training set is limited, and extrapolation to novel rotation axes is even more limited, implying that the networks do not infer full 3D structure, nor use linear interpolation. Yet, generalization is far superior to pure 2D matching. These findings help with designing datasets with 2D views required to achieve 3D generalization. Code to reproduce our experiments is publicly available: https://github.com/shoaibahmed/investigating_3d_generalization.git" @default.
- W4366552803 created "2023-04-22" @default.
- W4366552803 creator A5053635416 @default.
- W4366552803 creator A5054480988 @default.
- W4366552803 creator A5088840119 @default.
- W4366552803 date "2023-04-18" @default.
- W4366552803 modified "2023-10-16" @default.
- W4366552803 title "Investigating the Nature of 3D Generalization in Deep Neural Networks" @default.
- W4366552803 doi "https://doi.org/10.48550/arxiv.2304.09358" @default.
- W4366552803 hasPublicationYear "2023" @default.
- W4366552803 type Work @default.
- W4366552803 citedByCount "0" @default.
- W4366552803 crossrefType "posted-content" @default.
- W4366552803 hasAuthorship W4366552803A5053635416 @default.
- W4366552803 hasAuthorship W4366552803A5054480988 @default.
- W4366552803 hasAuthorship W4366552803A5088840119 @default.
- W4366552803 hasBestOaLocation W43665528031 @default.
- W4366552803 hasConcept C105795698 @default.
- W4366552803 hasConcept C108583219 @default.
- W4366552803 hasConcept C115961682 @default.
- W4366552803 hasConcept C119857082 @default.
- W4366552803 hasConcept C132459708 @default.
- W4366552803 hasConcept C134306372 @default.
- W4366552803 hasConcept C137800194 @default.
- W4366552803 hasConcept C153180895 @default.
- W4366552803 hasConcept C154945302 @default.
- W4366552803 hasConcept C162324750 @default.
- W4366552803 hasConcept C165064840 @default.
- W4366552803 hasConcept C177148314 @default.
- W4366552803 hasConcept C177264268 @default.
- W4366552803 hasConcept C187736073 @default.
- W4366552803 hasConcept C199360897 @default.
- W4366552803 hasConcept C2776760102 @default.
- W4366552803 hasConcept C2780451532 @default.
- W4366552803 hasConcept C2781238097 @default.
- W4366552803 hasConcept C33923547 @default.
- W4366552803 hasConcept C41008148 @default.
- W4366552803 hasConcept C50644808 @default.
- W4366552803 hasConcept C64876066 @default.
- W4366552803 hasConceptScore W4366552803C105795698 @default.
- W4366552803 hasConceptScore W4366552803C108583219 @default.
- W4366552803 hasConceptScore W4366552803C115961682 @default.
- W4366552803 hasConceptScore W4366552803C119857082 @default.
- W4366552803 hasConceptScore W4366552803C132459708 @default.
- W4366552803 hasConceptScore W4366552803C134306372 @default.
- W4366552803 hasConceptScore W4366552803C137800194 @default.
- W4366552803 hasConceptScore W4366552803C153180895 @default.
- W4366552803 hasConceptScore W4366552803C154945302 @default.
- W4366552803 hasConceptScore W4366552803C162324750 @default.
- W4366552803 hasConceptScore W4366552803C165064840 @default.
- W4366552803 hasConceptScore W4366552803C177148314 @default.
- W4366552803 hasConceptScore W4366552803C177264268 @default.
- W4366552803 hasConceptScore W4366552803C187736073 @default.
- W4366552803 hasConceptScore W4366552803C199360897 @default.
- W4366552803 hasConceptScore W4366552803C2776760102 @default.
- W4366552803 hasConceptScore W4366552803C2780451532 @default.
- W4366552803 hasConceptScore W4366552803C2781238097 @default.
- W4366552803 hasConceptScore W4366552803C33923547 @default.
- W4366552803 hasConceptScore W4366552803C41008148 @default.
- W4366552803 hasConceptScore W4366552803C50644808 @default.
- W4366552803 hasConceptScore W4366552803C64876066 @default.
- W4366552803 hasLocation W43665528031 @default.
- W4366552803 hasOpenAccess W4366552803 @default.
- W4366552803 hasPrimaryLocation W43665528031 @default.
- W4366552803 hasRelatedWork W2523475948 @default.
- W4366552803 hasRelatedWork W2536452361 @default.
- W4366552803 hasRelatedWork W2548164045 @default.
- W4366552803 hasRelatedWork W4223943233 @default.
- W4366552803 hasRelatedWork W4285183766 @default.
- W4366552803 hasRelatedWork W4294150926 @default.
- W4366552803 hasRelatedWork W4312200629 @default.
- W4366552803 hasRelatedWork W4318718858 @default.
- W4366552803 hasRelatedWork W4360585206 @default.
- W4366552803 hasRelatedWork W4364306694 @default.
- W4366552803 isParatext "false" @default.
- W4366552803 isRetracted "false" @default.
- W4366552803 workType "article" @default.