Researchers have developed EAST, a novel framework for early action prediction in videos. This strategy employs token masking to reduce memory usage and speed up training by twofold, with minimal impact on accuracy. EAST achieves state-of-the-art results on multiple benchmarks, including NTU60, SSv2, and UCF101, outperforming prior methods by significant margins. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
RANK_REASON The item describes a new academic paper detailing a novel framework and achieving state-of-the-art results on benchmarks.