Enhanced Motion-Text Alignment for Image-to-Video Transfer Learning
Open paper
Benchmark
Model
Rank
Results
action-classification-on-kinetics-400
MoTED (ViT-L/14, 32 frames)
–
Acc@1: 88.8
action-recognition-in-videos-on-something-something-v2
MoTED (ViT-L/14, 32 frames)
–
Top-1 Accuracy: 73.8
Rank counts only results with a code link.