AI Paper Atlas

Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning

Meta-Baseline: Exploring Simple Meta-Learning for Few-Shot Learning

Yinbo Chen, Zhuang Liu, Huijuan Xu, Trevor Darrell, Xiaolong Wang

ICCV 2021 · Label-efficient and robust learning › Zero-/few-shot and meta-learning · approach: Meta-learning analysis and benchmarks · 643 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 meta-learning stage over converged Classifier-BaselineClassifier-Baseline · miniImageNet 5-way 1-shotaccuracy (%): 58.91 -> 63.17Improved
add meta-learning stage over converged Classifier-BaselineClassifier-Baseline · tieredImageNet 5-way 5-shotaccuracy (%): 83.74 -> 83.74No clear effect
add meta-learning stage over converged Classifier-BaselineClassifier-Baseline (ResNet-50) · ImageNet-800 1-shotaccuracy (%): 86.07 -> 89.70Improved
add meta-learning stage over converged Classifier-BaselineClassifier-Baseline (ResNet-50) · ImageNet-800 5-shotaccuracy (%): 96.14 -> 96.14No clear effect
add whole-classification training stage before meta-learning (ablation)Meta-Baseline · miniImageNet 1-shot novel class generalizationaccuracy (%): 58.54 -> 63.33Improved
add whole-classification training stage before meta-learning (ablation)Meta-Baseline · miniImageNet 5-shot novel class generalizationaccuracy (%): 74.95 -> 80.02Improved
add whole-classification training stage before meta-learning (ablation)Meta-Baseline · miniImageNet 1-shot base class generalizationaccuracy (%): 86.74 -> 86.42Hurt
use cosine instead of squared Euclidean distance (ablation)Meta-Baseline · miniImageNet 5-way 1-shotaccuracy (%): 60.19 -> 63.33Improved
use cosine instead of squared Euclidean nearest-centroid at test time (ablation)Classifier-Baseline · miniImageNet 5-way 1-shotaccuracy (%): 56.29 -> 60.58Improved
shuffle classes so base and novel classes are closer (mini subset) (ablation)Meta-Baseline gain over Classifier-Baseline · tieredImageNet 1-shotaccuracy (%): +1.53 -> +4.24Improved
shuffle classes so base and novel classes are closer (full) (ablation)Meta-Baseline gain over Classifier-Baseline · tieredImageNet 1-shotaccuracy (%): +0.76 -> +2.81Improved
train classifier with cosine classifier instead of linear (ablation)Classifier-Baseline · miniImageNet 5-way 1-shotaccuracy (%): 60.58 -> 61.93Improved

Later papers that compared with this method

11 later accepted papers report Meta-baseline as a baseline in their results tables; 10 of them report a better number on most of the results they share with it.

3 more not listed.

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