ImageNet
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
ImageNet-1K 20% synthetic label noise (test)
60Accuracy
16
ImageNet 10% (train)
81.2Accuracy
16
ImageNet 1% (train)
75Accuracy
16
ImageNet 5-shot (train)
68.9Accuracy
16
ImageNet 2-shot (train)
62.3Accuracy
16
ImageNet 1-shot (train)
53.1Accuracy
16
ImageNet-100 20 tasks
66.36Average Accuracy
16
ImageNet 100 classes (test)
6.5FPR95
16
ImageNet 512x512 (train val)
2.41FID
16
ImageNet 100 (ID) OOD Average
5.76FPR95
16
ImageNet (val)
77.8Clean Accuracy
16
TieredImageNet
71.42Accuracy (0% Noise)
16
MiniImageNet
68.51Accuracy (0% Noise)
16
ImageNet (val)
5.2FID
16
ImageNet64
3.65BPD
16
miniImagenet
74.741-Shot Accuracy
16
ImageNet 100
89.1Seen Accuracy
16
ImageNet-Rendition (IN-R) (val)
92.5Top-1 Acc
16
ImageNet-9 Mixed-Next v1 (test)
55.5Accuracy
16
ImageNet-9 Mixed-Rand v1 (test)
58.2Accuracy
16
ImageNet-9 Mixed-Same v1 (test)
74.5Accuracy
16
ImageNet-9 Original v1 (test)
83.5Accuracy
16
ImageNet 1% labels 1k (val)
76.6Top-1 Accuracy
16
ImageNet 25% data
84.7IS
16
ImageNet 1k (val)
78.5Top-5 Acc (1%)
16