COCO
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
COCO 2017 (val)
70.93FID
12
COCO Robust-Seg 2017 (unseen datasets with synthetic degradations)
51.3AP
12
COCO Open-Vocabulary v2017
42.3AP50 (Novel)
12
COCO 2017
35.9APs
12
COCO-Position 2014
65.97AP
12
CorCOCO (val)
8.9D-ECE
12
COCO 2017 (val random split)
0.292Mask AP
12
COCO-Standard 2017 (10% labeled)
37.13mAP
12
COCO Standard 2017 (5% labeled)
33.01mAP
12
COCO Standard 1% labeled 2017
22.38mAP
12
COCO-standard 10% labeled 2017 (train)
35.08mAP
12
COCO novel and base categories 2014
37.4Novel AP50
12
COCO panoptic 133 categories (val)
52.7PQ
12
COCO-Crop (test)
80.69Accuracy
12
COCO-additional
50.45AP50:95
12
COCO Entities (test)
25.4BLEU-4
12
COCO mini 2017 (val)
35.6AP^mask
12
COCO-Stuff (test)
30.3Inception Score
12
COCO (val)
71.1AP
12
COCO Person category (val)
61.7AP
12
COCO (val)
93.59PCK@0.1
11
MS-COCO
44Runtime (s/img)
11
MS-COCO
23Runtime (s/img)
11
COCO 2014
56.52mAP
11
COCO subset (test)
59S2 Score
11