AI Paper Atlas

Research map / Label-efficient and robust learning / Self-supervised and contrastive

EfficientTrain: Exploring Generalized Curriculum Learning for Training Visual Backbones

Yulin Wang, Yang Yue, Rui Lu, Tianjiao Liu, Zhao Zhong, Shiji Song, Gao Huang

ICCV 2023 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Invariance-based self-supervised learning · 39 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
train with EfficientTrain (low-frequency cropping curriculum plus linear RandAug) instead of standard trainingvarious backbones · ImageNet-1Ktop-1 accuracy and training cost: 83.4% -> 83.6% (Swin-Base) at 1.50x speedupImproved
add linear RandAug on top of low-frequency cropping (ablation)EfficientTrain DeiT-Tiny · ImageNet-1Ktop-1 accuracy (%): 72.4 -> 73.3Improved
low-frequency cropping alone (ablation)baseline training · ImageNet-1Ktop-1 accuracy (%): 80.3 -> 80.0 (DeiT-S) at 1.5x speedupMixed
replace low-frequency cropping with image down-sampling (ablation)EfficientTrain · ImageNet-1Ktop-1 accuracy (%): 81.4 -> 81.0 (Swin-T)Hurt
use searched schedule of B instead of linear increasing (ablation)EfficientTrain · ImageNet-1Ktop-1 accuracy (%): 79.9 -> 80.4 (DeiT-S)Improved
shorter training schedule (100 epochs) with EfficientTrainDeiT-Small · ImageNet-1Ktop-1 accuracy (%): 75.5 -> 76.4Improved
use EfficientTrain with matched wall-time (450 epochs vs 300)DeiT-Tiny · ImageNet-1Ktop-1 accuracy (%): 72.5 -> 74.3Improved
increase final input size from 224 to 512EfficientTrain Swin-Base · ImageNet-1Ktraining speedup: 1.50x -> 3.37xImproved
pre-train MAE with EfficientTrainMAE ViT-B and ViT-L · ImageNet-1K fine-tuningtop-1 accuracy (%): ViT-B 83.5 -> 83.6; ViT-L 85.9 -> 85.8No clear effect
add FixRes on top of EfficientTrainCSWin-Small · ImageNet-1Ktop-1 accuracy (%): 83.6 -> 83.8Improved
pre-train detection backbones with EfficientTrainSwin backbones on COCO · COCO Cascade Mask-RCNNAP box: Swin-T 48.1 -> 48.2; Swin-S 50.0 -> 50.6No clear effect

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