CORe50
Benchmarks
TaskDatasetSOTA resultTrendResults
Core50 ID
98.9AUROC (COCO)
40
CORe50 (test)
96.7Test Accuracy
34
CORe50 indoor-to-outdoor sessions
84.5Accuracy
24
CORe50
97.1Average Accuracy
24
CORe50
90.6AVG Acc
22
CORe50 (test)
92.29A_T
22
CORe50 Hard
54.4s5 Score
18
CORe50 Unknown scenarios
94.37Average Accuracy (AA)
15
CORe50 5-shot
70.6mAP
13
CORe50 1-shot
60.9mAP
13
CORe50 multi-class 5-shot
99.3BCR
13
CORe50 Single-class five-shot, 1 novel class
99.8BCR
13
CORe50 (test)
99.1BCR
11
core50
61.37ACC
10
core50 Out-of-Sample
45.36ACC
9
core50 In-Sample
61.37Accuracy
9
CORe50 5 tasks
48.7mAP@50
8
CORe50 (instance)
71.45Accuracy
8
CoRe50 10 tasks, 5 classes per task
72.29Accuracy
7
CoRe50 5 tasks, 10 classes per task
72.96Accuracy
7
CORe50
96.87Average Accuracy
6
CORe50 (test)
0.48Latency (s)
4
CORe50
0.477Latency (ms)
3
Core50 (ID) vs ImageNet-1K (OOD) (Late split t=8)
7.73FPR95
2
CORe50
—Primary metric
0