Long Tail Learning On Inaturalist 2018

评估指标

Top-1 Accuracy

评测结果

各个模型在此基准测试上的表现结果

		Paper Title	Repository
LIFT (ViT-L/14@336px)	87.4%	Long-Tail Learning with Foundation Model: Heavy Fine-Tuning Hurts
LIFT (ViT-L/14)	85.2%	Long-Tail Learning with Foundation Model: Heavy Fine-Tuning Hurts
GML (ViT-B-16)	82.1%	Long-Tailed Recognition by Mutual Information Maximization between Latent Features and Ground-Truth Labels
VL-LTR (ViT-B-16)	81.0%	VL-LTR: Learning Class-wise Visual-Linguistic Representation for Long-Tailed Visual Recognition
LIFT (ViT-B/16)	80.4%	Long-Tail Learning with Foundation Model: Heavy Fine-Tuning Hurts
RAC (ViT-B-16)	80.24%	Retrieval Augmented Classification for Long-Tail Visual Recognition	-
GPaCo (2-R152)	79.8%	Generalized Parametric Contrastive Learning
GPaCo (ResNet-152)	78.1%	Generalized Parametric Contrastive Learning
TADE(ResNet-152)	77%	Self-Supervised Aggregation of Diverse Experts for Test-Agnostic Long-Tailed Recognition
ProCo (ResNet50)	75.8%	Probabilistic Contrastive Learning for Long-Tailed Visual Recognition
MDCS(Resnet50)	75.6%	MDCS: More Diverse Experts with Consistency Self-distillation for Long-tailed Recognition
GPaCo (ResNet-50)	75.4%	Generalized Parametric Contrastive Learning
CBD-ENS (ResNet-101)	75.3%	Class-Balanced Distillation for Long-Tailed Visual Recognition
PaCo(ResNet-152)	75.2%	Parametric Contrastive Learning
DeiT-LT	75.1%	DeiT-LT Distillation Strikes Back for Vision Transformer Training on Long-Tailed Datasets
APA (SE-ResNet-50)	74.8	Adaptive Parametric Activation
VL-LTR (ResNet-50)	74.6%	VL-LTR: Learning Class-wise Visual-Linguistic Representation for Long-Tailed Visual Recognition
GML (ResNet-50)	74.5%	Long-Tailed Recognition by Mutual Information Maximization between Latent Features and Ground-Truth Labels
NCL(ResNet-50)	74.2%	Nested Collaborative Learning for Long-Tailed Visual Recognition
BatchFormer(ResNet-50, RIDE)	74.1%	BatchFormer: Learning to Explore Sample Relationships for Robust Representation Learning

0 of 43 row(s) selected.