SCALE: Semantic- and Confidence-Aware Conditional Variational Autoencoder for Zero-shot Skeleton-based Action Recognition

Open paper
Benchmark Model Rank Results
zero-shot-skeletal-action-recognition-on-ntu-120SCALE–Accuracy (10 unseen classes): 73.6Accuracy (24 unseen classes): 54.9
zero-shot-skeletal-action-recognition-on-ntu-60SCALE–Accuracy (12 unseen classes): 50.0Accuracy (5 unseen classes): 84.5

Rank counts only results with a code link.