Action Detail Matters: Refining Video Recognition with Local Action Queries
Open paper
Benchmark
Model
Rank
Results
action-classification-on-kinetics-400
FocusVideo-L/14
–
Acc@1: 88.0
action-recognition-in-videos-on-something-something-v2
FocusVideo-L/14
–
Top-1 Accuracy: 72.9
Rank counts only results with a code link.