Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

USIS

Benchmarks

Task NameDataset NameSOTA ResultTrend
Open-vocabulary SegmentationUSIS16K
Average Accuracy (aAcc)60.45
18
Open-vocabulary SegmentationUSIS10K
aAcc75.04
18
Multi-class Instance SegmentationUSIS10K
mAP66.8
16
Instance SegmentationUSIS 16K
mAP85.7
11
Class-Agnostic Instance SegmentationUSIS10K
mAP64.6
11
Salient Instance SegmentationUSIS16K source (test)
mAP85.7
9
Showing 6 of 6 rows