ImageNet
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
ImageNet and OOD variants (ImV2, Im-A, Im-R, Im-S) 1.0 (val)
0.842ImNet Acc
8
ImageNet-1M 1% labeled
70.3NMI
8
ImageNet Hue-shifted (test)
68.74Accuracy
8
ImageNet Noise
0.15Classification Error
8
ImageNet 1000 random images
0.246C3 Truthful Gap
8
ImageNet 1k (0.20%)
48.6Accuracy
8
ImageNet (test)
70Human Preference Count
8
ImageNet (test)
0.027CWCE
8
ImageNet-C mean over shifts v1 (test)
83.7Accuracy
8
ImageNet 1k (10%)
71.2Top-1 Acc
8
ImageNet
4.3ECE
8
ImageNet
1.65FID
8
ImageNetV2 Top Images Ti (test)
79.4Top-1 Acc
8
ImageNetV2 Threshold 0.7 (test)
74.5Top-1 Acc
8
ImageNet (val)
42.05FLOPs (%)
8
ImageNet-200
61.03Mf
8
ImageNet 100
63Mf Score
8
ImageNet 500 base + 10x50 novel classes (incremental)
70.67Accuracy (t=0..10)
8
Red Mini-ImageNet 0.6 noise rate Controlled Noisy Web Labels (CNWL) (test)
53.21Accuracy
8
ImageNet (All)
3Top-1 Accuracy
8
ImageNet Superclasses (591 classes) WordNet hierarchy (val)
46.7Accuracy
8
ImageNet Superclasses 466 classes WordNet hierarchy (val)
60Accuracy
8
ImageNet Superclasses 128 classes WordNet hierarchy (val)
0.718Accuracy
8
ImageNet Superclasses (29 classes) WordNet hierarchy (val)
79.7Accuracy
8
ImageNet Superclasses (10 classes) WordNet hierarchy (val)
85.7Accuracy
8