Work

Optimized Temporal Convolutional Networks with Multi-Stage Attention for Efficient Human Action Recognition