AI Paper Atlas

Research map / Detection and segmentation / Open-set and few-shot detection

EVA: Exploring the Limits of Masked Visual Representation Learning at Scale

Yuxin Fang, Wen Wang, Binhui Xie, Quan Sun, Ledell Wu, Xinggang Wang, Tiejun Huang, Xinlong Wang, Yue Cao

CVPR 2023 · Detection and segmentation › Open-set and few-shot detection · approach: Open-vocabulary 3D and world perception · 478 citations (OpenAlex)

Read the paper

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
Use masked CLIP-feature reconstruction without additional feature tokenization rather than tokenization-based masked modeling (ablation)ViT-B pre-training · ImageNet-1Ktop-1 accuracy: 85.0 -> 85.0No clear effect
Use masked CLIP-feature reconstruction without additional feature tokenization rather than tokenization-based masked modeling (ablation)ViT-B pre-training · ADE20Ksingle-scale mIoU: 52.6 -> 52.7No clear effect
Extend tokenized feature pre-training from 300 to 1600 epochs (ablation)ViT-B pre-training · ImageNet-1Ktop-1 accuracy: 85.0 -> 85.5Improved
Extend tokenized feature pre-training from 300 to 1600 epochs (ablation)ViT-B pre-training · ADE20Ksingle-scale mIoU: 52.7 -> 53.1Improved
Add feature distillation during 300-epoch pre-training (ablation)ViT-B pre-training · ImageNet-1Ktop-1 accuracy: 85.0 -> 85.1No clear effect
Add feature distillation during 300-epoch pre-training (ablation)ViT-B pre-training · ADE20Ksingle-scale mIoU: 52.6 -> 52.5Hurt
Extend feature-distillation pre-training from 300 to 800 epochs (ablation)ViT-B pre-training · ImageNet-1Ktop-1 accuracy: 85.1 -> 85.1No clear effect
Extend feature-distillation pre-training from 300 to 800 epochs (ablation)ViT-B pre-training · ADE20Ksingle-scale mIoU: 52.5 -> 52.7Improved
Use masked CLIP-feature reconstruction instead of 800-epoch feature distillation (ablation)ViT-B pre-training · ImageNet-1Ktop-1 accuracy: 85.1 -> 85.5Improved
Use masked CLIP-feature reconstruction instead of 800-epoch feature distillation (ablation)ViT-B pre-training · ADE20Ksingle-scale mIoU: 52.7 -> 53.3Improved
Initialize the CLIP vision tower from pre-trained EVA rather than training it from scratch1.1B CLIP-g · zero-shot classification, 12-benchmark averagetop-1 accuracy: +3.8Improved
Use EVA as the CLIP vision tower initialization1.1B CLIP-g · zero-shot classification, natural-distribution-shift benchmarksaccuracy gap: 2.5Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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