AIM: Adapting Image Models for Efficient Video Action Recognition

Benchmark Model Rank Results
action-recognition-on-diving-48AIM (CLIP ViT-L/14, 32x224)#2Accuracy: 90.6
action-classification-on-kinetics-700AIM (CLIP ViT-L/14, 32x224)#14Top-1 Accuracy: 80.4
action-classification-on-kinetics-400AIM (CLIP ViT-L/14, 32x224)#32Acc@1: 87.5Acc@5: 97.7