Gated-temporal Attention

The GLSFormer is a transformer-based model for automated surgical step recognition. It seeks to address limitations of current state-of-the-art methods which either model spatial and temporal information separately or focus…