MS-COCO
Benchmarks
TaskDatasetSOTA resultTrendResults
MS-COCO 10% labeled protocol
34.82mAP
19
MS-COCO (test)
89.5Classification Accuracy
18
MS-COCO Random Setting N=4 (test)
98.66Accuracy
18
MS-COCO 1K
29.736HPSv2.1
18
MS-COCO
83.3mAP @ p=0.1
18
MS-COCO 512x512 with SDv3.5
82.68FID
18
MS-COCO 30-shot (test)
34.9AP
18
MS-COCO (val)
81.5F1 Score (All)
18
MS COCO (val)
51AP
18
MS-COCO 2014 (val)
78mAP (Overall)
18
MS-COCO English In-domain (test)
44.1BLEU-4
17
MS-COCO 2014
26.93CLIP Score
16
MS COCO
112.3CIDEr
16
MS-COCO (test)
100Detection Rate
16
MS-COCO 200 images (test)
1,020.38Tokens
16
MS COCO
0.862Insertion Score
16
MS-COCO
4.92FID-30K
16
MS COCO
72.5mAP (10% known)
16
MS COCO 2014 2017 (val)
52.9AP (Box)
16
MS-COCO 48/17
61.7Recall@100 (IoU=0.5)
16
MS-COCO (65/15)
40.55Seen mAP
16
MS-COCO Random
94.92Accuracy
15
MS-COCO FLUX.1-dev (test)
22.58FID
15
MS-COCO
91.6Accuracy (Rand.)
15
MS COCO 2017
86.4mAP
15