Spyros Gidaris

DINO-Foresight forecasts future VFM features to enable robust and scalable semantic scene understanding across multiple tasks.

Jointly modeling image laments and high-level semantic features (e.g. DINO), significantly enhances generative quality and training efficiency.