DUGO: Dual-Scale Gaussian Memory for Online Temporal Action Localization in Egocentric Procedural Videos
DUGO (Dual-Scale Encoding and Gaussian Memory Online), a new OnTAL framework that jointly models local motion dynamics and long-range temporal dependencies, uses a Dual-Scale Temporal Encoder and introduces a Gaussian Memory Bank, which represents action memory as stochastic, input-conditioned Gaussian prototypes.