AI Paper Atlas

Research map / Vision-language and multimodal / Cross-modal learning

Omnivore: A Single Model for Many Visual Modalities

Rohit Girdhar, Mannat Singh, Nikhila Ravi, Laurens van der Maaten, Armand Joulin, Ishan Misra

CVPR 2022 · Vision-language and multimodal › Cross-modal learning · approach: Multi-modality image fusion · 196 citations (OpenAlex)

Check your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Jointly train one shared model on image, video, and RGBD datasets instead of modality-specific modelsSwin-B · Kinetics-400top-1 accuracy: 80.6 -> 83.3Improved
Jointly train one shared model on image, video, and RGBD datasets instead of modality-specific modelsSwin-B · SUN RGB-Dtop-1 accuracy: 64.8 -> 65.4Improved
Jointly train one shared model on image, video, and RGBD datasets instead of modality-specific modelsSwin-L · Kinetics-400top-1 accuracy: 83.1 -> 84.1Improved
Jointly train one shared model on image, video, and RGBD datasets instead of modality-specific modelsSwin-L · SUN RGB-Dtop-1 accuracy: 69.1 -> 67.1Hurt
Use an Omnivore pretrained representation for downstream finetuningSwin-L · iNaturalist-2018classification accuracy: 81.7 -> 84.1Improved
Use an Omnivore pretrained representation for downstream finetuningSwin-B · Something Something-v2top-1 accuracy: +0.5Improved
Initialize from ImageNet-21K and jointly finetune instead of training from scratch (ablation)Omnivore · ImageNet-1Ktop-1 accuracy: --0.7Hurt
Initialize from ImageNet-21K and jointly finetune instead of training from scratch (ablation)Omnivore · Kinetics-400top-1 accuracy: --0.9Hurt
Initialize from ImageNet-21K and jointly finetune instead of training from scratch (ablation)Omnivore · SUN RGB-Dtop-1 accuracy: +0.9Improved
Increase SUN RGB-D replication from 50 to 100 (ablation)Omnivore · SUN RGB-Dtop-1 accuracy change: --2.1Hurt
Increase ImageNet-21K proportion from 0.1 to 1.0 (ablation)Omnivore · Kinetics-400top-1 accuracy change: --4.5Hurt
Increase ImageNet-21K proportion from 0.1 to 1.0 (ablation)Omnivore · SUN RGB-Dtop-1 accuracy change: +2.0Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy