Gen2Balance: Generative Video Balances Long-Tailed Action Recognition at ECCV 2026

A paper from the University of Bristol and Adobe Research, presenting at ECCV 2026, uses text-to-video generation to fill gaps in long-tailed video action recognition datasets, achieving SOTA results and releasing a dataset with 223 classes.

2026-09-10 ~ 2026-09-10 · 2 related posts

• Gen2Balance: text-to-video synthesis fills long-tailed data for SOTA action recognition — dimadamen · 2026-09-10
• Gen2Balance at ECCV 2026: generative videos fill long-tailed action datasets to reach SOTA — dimadamen · 2026-09-10