HT-AggNet: Hierarchical temporal aggregation network with near-zero-cost layer stacking for human activity recognition

Jaegyun Park, Dae-Won Kim, Jaesung Lee

Abstract: Highlights•A hierarchical temporal aggregation network (HT-AggNet) is proposed.•Temporal glance convolution to model global context information of signal patterns.•Near-zero-cost layer stacking with negligible increase in computational costs.•HT-AggNet hierarchically extracts local and global temporal information.•HT-AggNet outperforms existing methods on seven publicly available datasets.