Action100M: A Large-scale Video Action Dataset
Open paper
Benchmark
Model
Rank
Results
zero-shot-action-recognition-on-kinetics-400-zero-shot-action-classification
VL-JEPA Stage 3 ViT-L/256 (32 frames, zero-shot) (extra data)
#3
Accuracy: 64.8
Rank counts only results with a code link.