List your company
ResearchFormerly WizWand

Semi-supervised Image Classification on ImageNet 1k (val)

78.5Top-5 Acc (1%)SwAV
19.7434.99550.2565.505May 11, 2020
Updated 8mo ago

Evaluation results

MethodLinks
SwAVarchitecture=ResNet-50-MLP, #pretrain epochs=800, batch size=4096, multi-crop augmentation=true2020.0578.589.9——
BYOLarchitecture=ResNet-50-MLPbig, #pretrain epochs=1000, batch size=40962020.0578.489——
SimCLRarchitecture=ResNet-50-MLP, #pretrain epochs=1000, batch size=40962020.0575.587.8——
PCLarchitecture=ResNet-50, #pretrain epochs=2002020.0575.385.6——
PCL v2architecture=ResNet-50-MLP, #pretrain epochs=2002020.0573.985——
MoCo v2architecture=ResNet-50-MLP, #pretrain epochs=2002020.0566.384.4——
PIRLarchitecture=ResNet-50, #pretrain epochs=8002020.0557.283.8——
MoCoarchitecture=ResNet-50, #pretrain epochs=2002020.0556.983——
SimCLRarchitecture=ResNet-50-MLP, #pretrain epochs=2002020.0556.582.7——
S^4L Rotationarchitecture=ResNet-50v2, #pretrain epochs=-2020.0553.483.8——
Pseudolabelsarchitecture=ResNet-50v2, #pretrain epochs=-2020.0551.682.4——
Supervised baselinearchitecture=ResNet-50, #pretrain epochs=-2020.0548.480.4——
VAT + Entropy Min.architecture=ResNet-50v2, #pretrain epochs=-2020.054783.4——
Jigsawarchitecture=ResNet-50, #pretrain epochs=902020.0545.379.3——
Instance Discriminationarchitecture=ResNet-50, #pretrain epochs=2002020.0539.277.4——
Randomarchitecture=ResNet-50, #pretrain epochs=-2020.052259——
CoMatchLabeled data ratio=10%2024.10——73.691.6
SimMatchLabeled data ratio=10%2024.10——74.191.5
UniMatch V2Labeled data ratio=10%2024.10——74.391.7