Matches in SemOpenAlex for { <https://semopenalex.org/work/W4382771804> ?p ?o ?g. }
Showing items 1 to 88 of
88
with 100 items per page.
- W4382771804 abstract "Dense pixel-specific representation learning at scale has been bottlenecked due to the unavailability of large-scale multi-view datasets. Current methods for building effective pretraining datasets heavily rely on annotated 3D meshes, point clouds, and camera parameters from simulated environments, preventing them from building datasets from real-world data sources where such metadata is lacking. We propose a pretraining dataset-curation approach that does not require any additional annotations. Our method allows us to generate multi-view datasets from both real-world videos and simulated environments at scale. Specifically, we experiment with two scales: MIMIC-1M with 1.3M and MIMIC-3M with 3.1M multi-view image pairs. We train multiple models with different masked image modeling objectives to showcase the following findings: Representations trained on our automatically generated MIMIC-3M outperform those learned from expensive crowdsourced datasets (ImageNet-1K) and those learned from synthetic environments (MULTIVIEW-HABITAT) on two dense geometric tasks: depth estimation on NYUv2 (1.7%), and surface normals estimation on Taskonomy (2.05%). For dense tasks which also require object understanding, we outperform MULTIVIEW-HABITAT, on semantic segmentation on ADE20K (3.89%), pose estimation on MSCOCO (9.4%), and reduce the gap with models pre-trained on the object-centric expensive ImageNet-1K. We outperform even when the representations are frozen, and when downstream training data is limited to few-shot. Larger dataset (MIMIC-3M) significantly improves performance, which is promising since our curation method can arbitrarily scale to produce even larger datasets. MIMIC code, dataset, and pretrained models are open-sourced at https://github.com/RAIVNLab/MIMIC." @default.
- W4382771804 created "2023-07-02" @default.
- W4382771804 creator A5013406069 @default.
- W4382771804 creator A5018421415 @default.
- W4382771804 creator A5019820384 @default.
- W4382771804 creator A5025377024 @default.
- W4382771804 creator A5030320952 @default.
- W4382771804 creator A5032451496 @default.
- W4382771804 creator A5076121553 @default.
- W4382771804 date "2023-06-26" @default.
- W4382771804 modified "2023-10-12" @default.
- W4382771804 title "MIMIC: Masked Image Modeling with Image Correspondences" @default.
- W4382771804 doi "https://doi.org/10.48550/arxiv.2306.15128" @default.
- W4382771804 hasPublicationYear "2023" @default.
- W4382771804 type Work @default.
- W4382771804 citedByCount "0" @default.
- W4382771804 crossrefType "posted-content" @default.
- W4382771804 hasAuthorship W4382771804A5013406069 @default.
- W4382771804 hasAuthorship W4382771804A5018421415 @default.
- W4382771804 hasAuthorship W4382771804A5019820384 @default.
- W4382771804 hasAuthorship W4382771804A5025377024 @default.
- W4382771804 hasAuthorship W4382771804A5030320952 @default.
- W4382771804 hasAuthorship W4382771804A5032451496 @default.
- W4382771804 hasAuthorship W4382771804A5076121553 @default.
- W4382771804 hasBestOaLocation W43827718041 @default.
- W4382771804 hasConcept C111919701 @default.
- W4382771804 hasConcept C115961682 @default.
- W4382771804 hasConcept C119857082 @default.
- W4382771804 hasConcept C121332964 @default.
- W4382771804 hasConcept C121684516 @default.
- W4382771804 hasConcept C127413603 @default.
- W4382771804 hasConcept C131979681 @default.
- W4382771804 hasConcept C153180895 @default.
- W4382771804 hasConcept C154945302 @default.
- W4382771804 hasConcept C17744445 @default.
- W4382771804 hasConcept C199539241 @default.
- W4382771804 hasConcept C200601418 @default.
- W4382771804 hasConcept C2776359362 @default.
- W4382771804 hasConcept C2778755073 @default.
- W4382771804 hasConcept C2780505938 @default.
- W4382771804 hasConcept C2781238097 @default.
- W4382771804 hasConcept C31487907 @default.
- W4382771804 hasConcept C31972630 @default.
- W4382771804 hasConcept C41008148 @default.
- W4382771804 hasConcept C62520636 @default.
- W4382771804 hasConcept C89600930 @default.
- W4382771804 hasConcept C93518851 @default.
- W4382771804 hasConcept C94625758 @default.
- W4382771804 hasConceptScore W4382771804C111919701 @default.
- W4382771804 hasConceptScore W4382771804C115961682 @default.
- W4382771804 hasConceptScore W4382771804C119857082 @default.
- W4382771804 hasConceptScore W4382771804C121332964 @default.
- W4382771804 hasConceptScore W4382771804C121684516 @default.
- W4382771804 hasConceptScore W4382771804C127413603 @default.
- W4382771804 hasConceptScore W4382771804C131979681 @default.
- W4382771804 hasConceptScore W4382771804C153180895 @default.
- W4382771804 hasConceptScore W4382771804C154945302 @default.
- W4382771804 hasConceptScore W4382771804C17744445 @default.
- W4382771804 hasConceptScore W4382771804C199539241 @default.
- W4382771804 hasConceptScore W4382771804C200601418 @default.
- W4382771804 hasConceptScore W4382771804C2776359362 @default.
- W4382771804 hasConceptScore W4382771804C2778755073 @default.
- W4382771804 hasConceptScore W4382771804C2780505938 @default.
- W4382771804 hasConceptScore W4382771804C2781238097 @default.
- W4382771804 hasConceptScore W4382771804C31487907 @default.
- W4382771804 hasConceptScore W4382771804C31972630 @default.
- W4382771804 hasConceptScore W4382771804C41008148 @default.
- W4382771804 hasConceptScore W4382771804C62520636 @default.
- W4382771804 hasConceptScore W4382771804C89600930 @default.
- W4382771804 hasConceptScore W4382771804C93518851 @default.
- W4382771804 hasConceptScore W4382771804C94625758 @default.
- W4382771804 hasLocation W43827718041 @default.
- W4382771804 hasLocation W43827718042 @default.
- W4382771804 hasOpenAccess W4382771804 @default.
- W4382771804 hasPrimaryLocation W43827718041 @default.
- W4382771804 hasRelatedWork W1482423459 @default.
- W4382771804 hasRelatedWork W1789078735 @default.
- W4382771804 hasRelatedWork W2026539069 @default.
- W4382771804 hasRelatedWork W207884067 @default.
- W4382771804 hasRelatedWork W2110352794 @default.
- W4382771804 hasRelatedWork W2129918226 @default.
- W4382771804 hasRelatedWork W2365973415 @default.
- W4382771804 hasRelatedWork W2996457675 @default.
- W4382771804 hasRelatedWork W3127016596 @default.
- W4382771804 hasRelatedWork W4237235066 @default.
- W4382771804 isParatext "false" @default.
- W4382771804 isRetracted "false" @default.
- W4382771804 workType "article" @default.