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

MeViS

Benchmarks

Task NameDataset NameSOTA ResultTrend
Referring Video Object SegmentationMeViS (val)
J&F Score0.635
161
Referring Video SegmentationMeViS
J&F Score73.1
81
Referring Video SegmentationMeViS (test)
J&F Score53.7
25
Video Referring Expression SegmentationMeViS (val-u)
J&F Score70.8
18
Referring Video Object SegmentationMeViS (val-u)
J&F Score72.2
17
Motion localizationMeViS
SL Score68
15
Referring Video Object SegmentationMeViS v2 (val)
J&F43.9
8
Referring Video Object SegmentationMeViS v1 (val)
J&F Score47.6
8
Referring Video Object SegmentationMeViS-Text PVUW 2026 (test)
J&F Score78.97
6
Audio-Guided Video Object SegmentationMeViS v2
J&F Score42.3
6
Video Object GroundingMeViS
J Score62.3
6
Text-to-Video RetrievalMeViS
Recall@155.6
6
Video-to-Text RetrievalMeViS
R@1 (V2T)58.4
6
Video-to-Text RetrievalMeViS (test)
R@159.2
5
Text-to-Video RetrievalMeViS (test)
R@10.568
5
Trajectory GenerationMeViS (test)
AJ0.28
5
Text-to-Trajectory RetrievalMeViS
Recall@134.2
5
Video Reasoning SegmentationMeViS
FPS5.1
4
Referring Multi-Object TrackingMeViS v2
HOTA*38.8
4
Referring Motion Expression GenerationMeViS
METEOR15.68
4
Object-centric motion localizationMeViS non-static clips
SL Score53
3
Spatial GroundingMeViS (val)
J Score62.3
3
Referring Multi-Object TrackingMeViS v2 (test)
HOTA*-
0
Audio-Guided Video Object SegmentationMeViS v2 (test)
J&F Score-
0
Showing 24 of 24 rows