List your company
ResearchFormerly WizWand

Few-shot Classification on ChestX

26.05AccuracyPsCo
22.118823.139424.1625.1806Mar 2, 2023Apr 10, 2023May 19, 2023Jun 27, 2023Aug 5, 2023Sep 13, 2023Oct 23, 2023
Updated 8mo ago

Evaluation results

MethodLinks
PsCoBackbone=ResNet-50, Base Method=BYOL, Evaluation protocol=5-way 5-shot2023.0326.05——
CLIPNumber of ways=5, Number of shots=52023.1025.58——
SupervisedBackbone=ResNet-50, Evaluation protocol=5-way 5-shot2023.0325.35——
WiSE-FTNumber of ways=5, Number of shots=52023.1025.08——
PsCoBackbone=ResNet-18, Base Method=MoCo v2, Evaluation protocol=5-way 5-shot2023.0325.03——
FD-AlignNumber of ways=5, Number of shots=52023.1024.95——
MoCo v2Backbone=ResNet-18, Evaluation protocol=5-way 5-shot2023.0324.34——
MoCo v2Backbone=ResNet-50, Evaluation protocol=5-way 5-shot2023.0324.28——
BYOLBackbone=ResNet-50, Evaluation protocol=5-way 5-shot2023.0324.15——
PsCoBackbone=ResNet-50, Base Method=MoCo v2, Evaluation protocol=5-way 5-shot2023.0323.6——
CLIPNumber of ways=5, Number of shots=12023.1022.65——
FD-AlignNumber of ways=5, Number of shots=12023.1022.31——
WiSE-FTNumber of ways=5, Number of shots=12023.1022.27——
MAMLLearning Paradigm=Supervised meta-learning, Backbone=Conv5, Meta-training dataset=miniImageNet, Number of evaluation tasks=20002023.03—22.6124.25
Meta-GMVAELearning Paradigm=Unsupervised meta-learning, Backbone=Conv5, Meta-training dataset=miniImageNet, Number of evaluation tasks=20002023.03—23.2326.26
Meta-SVEBMLearning Paradigm=Unsupervised meta-learning, Backbone=Conv5, Meta-training dataset=miniImageNet, Number of evaluation tasks=2002023.03—26.2628.91
MoCo v2Learning Paradigm=Self-supervised learning, Backbone=Conv5, Meta-training dataset=miniImageNet, Number of evaluation tasks=20002023.03—25.2429.19
ProtoNetsLearning Paradigm=Supervised meta-learning, Backbone=Conv5, Meta-training dataset=miniImageNet, Number of evaluation tasks=20002023.03—23.1525.72
PsCoLearning Paradigm=Unsupervised meta-learning, Backbone=Conv5, Meta-training dataset=miniImageNet, Number of evaluation tasks=20002023.03—24.7827.69
SimCLRLearning Paradigm=Self-supervised learning, Backbone=Conv5, Meta-training dataset=miniImageNet, Number of evaluation tasks=20002023.03—25.1429.21
SwAVLearning Paradigm=Self-supervised learning, Backbone=Conv5, Meta-training dataset=miniImageNet, Number of evaluation tasks=20002023.03—24.9928.57