List your company
ResearchFormerly WizWand

Fine-grained Image Classification on FGVC (CUB-200, Stanford Dogs, Stanford Cars, NABirds) (test)

90.06Average Top-1 AccuracyLoCA (r16)
43.603255.664167.72579.7859Jul 8, 2026
Updated 4mo ago

Evaluation results

MethodLinks
LoCA (r16)Backbone=ConvNeXt-B, Number of Trainable Parameters (M)=3.702026.0790.06
CABackbone=ConvNeXt-B, Number of Trainable Parameters (M)=6.132026.0789.28
LoRABackbone=ConvNeXt-B, Number of Trainable Parameters (M)=17.582026.0788.18
FSFBackbone=ConvNeXt-B, Number of Trainable Parameters (M)=1.112026.0788.04
CoLoRABackbone=ConvNeXt-B, Number of Trainable Parameters (M)=4.572026.0786.11
LoCA (r16)Backbone=ResNet-50, Number of Trainable Parameters (M)=1.942026.0783.67
CABackbone=ResNet-50, Number of Trainable Parameters (M)=2.232026.0783.48
FSFBackbone=ResNet-50, Number of Trainable Parameters (M)=2.522026.0781.07
LoRABackbone=ResNet-50, Number of Trainable Parameters (M)=2.432026.0780.96
FFTBackbone=ConvNeXt-B, Number of Trainable Parameters (M)=87.872026.0779.73
LPBackbone=ConvNeXt-B, Number of Trainable Parameters (M)=0.312026.0777.55
CoLoRABackbone=ResNet-50, Number of Trainable Parameters (M)=0.992026.0776.44
FFTBackbone=ResNet-50, Number of Trainable Parameters (M)=24.142026.0775.73
BiasBackbone=ConvNeXt-B, Number of Trainable Parameters (M)=0.442026.0764.98
BiasBackbone=ResNet-50, Number of Trainable Parameters (M)=0.672026.0748.85
LPBackbone=ResNet-50, Number of Trainable Parameters (M)=0.622026.0745.39