ImageNet-R
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
ImageNet-R B0 Inc20
83.87Last Accuracy
107
ImageNet-R 10 tasks
87.18Average ACC@10
64
ImageNet-R 1.0 (val)
71.1Top-1 Acc
58
Split ImageNet-R
1.27Average Forgetting Measure
57
ImageNet-R 10-task
82.06FAA
54
Split ImageNet-R 10 incremental tasks
82.38Class Accuracy
40
ImageNet-R (20 tasks)
86.26Average Accuracy (20 Tasks)
38
ImageNet-R 10-task split
82.06FAA
26
ImageNet-R OfficeHome source (test)
0.772Accuracy
25
ImageNet-R MiniDomainnet source (test)
70.18Accuracy
25
ImageNet-R TI2I
31.57CLIP Score
24
ImageNet-R B0 Inc20 (test)
85.63Average Performance (A-bar)
23
ImageNet-R (10-Split)
91.94Accuracy
22
ImageNet-R 20-split
79.4Accuracy
21
ImageNet-R (test)
87.55Accuracy
20
ImageNet-R Si-Blurry
51.86AAUC
18
ImageNet-R (20-split)
34.8Accuracy
16
ImageNet-R (10-split)
54.48Accuracy (ACC)
16
ImageNet-R 5-split
55.65Accuracy
16
ImageNet-R N = 10
80.64Accuracy
16
ImageNet-R 20 tasks rho=0.01
70Accuracy (Last Task)
14
ImageNet-R 20 tasks rho=0.1
73.7Accuracy (Last Task)
14
ImageNet-R strong appearance-shift (test)
44.2Accuracy
14
ImageNet-R 10 tasks 1.0 (test)
71.84ACC10
14
ImageNet-R 10 sequential tasks 200 classes
82.92A1:N
14