Semi Supervised Image Classification On 2

评估指标

Top 1 Accuracy

Top 5 Accuracy

评测结果

各个模型在此基准测试上的表现结果

			Paper Title
Semi-ViT (ViT-Huge)	84.3%	96.6%	Semi-supervised Vision Transformers at Scale
SimCLRv2 self-distilled (ResNet-152 x3, SK)	80.9%	95.5%	Big Self-Supervised Models are Strong Semi-Supervised Learners
SimCLRv2 (ResNet-152 x3, SK)	80.1%	95.0%	Big Self-Supervised Models are Strong Semi-Supervised Learners
SimCLRv2 distilled (ResNet-50 x2, SK)	80.2%	95.0%	Big Self-Supervised Models are Strong Semi-Supervised Learners
SimCLRv2 distilled (ResNet-50)	77.5%	93.4%	Big Self-Supervised Models are Strong Semi-Supervised Learners
TWIST (ResNet-50 x2)	75.3%	92.8%	Self-Supervised Learning by Estimating Twin Class Distributions
SimMatch + EPASS (ResNet-50)	75.3%	92.6	Debiasing, calibrating, and improving Semi-supervised Learning performance via simple Ensemble Projector
SimCLR (ResNet-50 4×)	-	92.6%	A Simple Framework for Contrastive Learning of Visual Representations
SequenceMatch (ResNet-50)	75.2%	91.9	SequenceMatch: Revisiting the design of weak-strong augmentations for Semi-supervised learning
SimCLRv2 (ResNet-50 x2)	73.9%	91.9%	Big Self-Supervised Models are Strong Semi-Supervised Learners
CoMatch + EPASS (ResNet-50)	74.1%	91.5	Debiasing, calibrating, and improving Semi-supervised Learning performance via simple Ensemble Projector
CoMatch (w. MoCo v2)	73.7%	91.4%	CoMatch: Semi-supervised Learning with Contrastive Graph Regularization
Meta Pseudo Labels (ResNet-50)	73.89%	91.38%	Meta Pseudo Labels
CowMix (ResNet-152)	73.94%	91.24%	Milking CowMask for Semi-Supervised Image Classification
S4L-MOAM (ResNet-50 4×)	73.21%	91.23%	S4L: Self-Supervised Semi-Supervised Learning
Rotation + VAT + Ent. Min.	-	91.23%	S4L: Self-Supervised Semi-Supervised Learning
WCL (ResNet-50)	72.0%	91.2%	Weakly Supervised Contrastive Learning
SimCLR (ResNet-50 2×)	-	91.2%	A Simple Framework for Contrastive Learning of Visual Representations
CPC v2 (ResNet-161)	73.1%	91.2%	Data-Efficient Image Recognition with Contrastive Predictive Coding
RELICv2 (ResNet-50)	72.4%	91.2%	Pushing the limits of self-supervised ResNets: Can we outperform supervised learning without labels on ImageNet?

0 of 68 row(s) selected.