AI Paper Atlas

Research map / Vision-language and multimodal / CLIP and zero-shot

The Effectiveness of MAE Pre-Pretraining for Billion-Scale Pretraining

Mannat Singh, Quentin Duval, Kalyan Vasudev Alwala, Haoqi Fan, Vaibhav Aggarwal, Aaron Adcock, Armand Joulin, Piotr Dollar, Christoph Feichtenhofer, Ross Girshick, Rohit Girdhar, Ishan Misra

ICCV 2023 · Vision-language and multimodal › CLIP and zero-shot · approach: Adapting VLMs for zero-shot classification · 74 citations (OpenAlex)

Read the paperCheck 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
add MAE pre-pretraining before weakly supervised pretraining (ablation)WSP on IG-3B · ImageNet-1k linear probetop-1 accuracyImproved
add MAE pre-pretraining (2B vs 6.5B without) (ablation)WSP · ImageNet-1k linear probetop-1 accuracyImproved
increase pre-pretraining MAE epochs beyond 1 (ablation)MAE->WSP · IG-3B pretrainingtransfer accuracyNo clear effect
add MAE pre-pretraining under fixed training FLOPs (ablation)WSP · IG-3B pretrainingtransfer performance: up to 2x more efficientImproved
pre-pretrain MAE on ImageNet-1k instead of IG-3B (ablation)MAE->WSP ViT-H · ImageNet-1k finetunetop-1 accuracy: 89.3 -> 89.4No clear effect
add MAE pre-pretraining to WSP for object detection scaling (ablation)WSP ViT · LVIS/COCO detectionAPImproved
add MAE pre-pretraining on ImageNet-21k pretraining (ablation)WSP ViT-L · IN21k pretrainingtransfer accuracyImproved
scale MAE pretraining data from IN1k to IG-3B (ablation)MAE · downstream tasksaccuracyImproved
MAE pre-pretraining (MAWS ViT-H) vs MAE on IN1k for COCO detectionMAWS ViT-H · COCO val2017AP box: 58.7 -> 57.7Hurt
add MAE pre-pretraining to SWAG-style WSP ViT-H on LVISWSP ViT-H · LVIS v1 valAP box: 47.1 -> 50.8Improved
add relative position embeddings in video finetuning (ablation)MAWS ViT-L · Kinetics-400top-1 accuracy: +0.6Improved
scale WSP training hyperparameters choices (loss, head, augmentation, lr) (ablation)WSP · IG-3B 1 epochaccuracyNo clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 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