A frame selection and spatiotemporal attention incorporated video transformer for third-view industrial assembly action recognition | Synapse