The authors freeze a pretrained static novel-view synthesis model and attach a lightweight dynamic encoder.
A static scene model learns motion to render new viewpoints
FL4N adds a trainable temporal branch to a frozen view-synthesis model, reporting strong results across eight synthetic and real-world benchmarks.
Big Tech
Eunbeen Hong · Hyeonseo Yu · Honggyu An · Jooho Son · Hyunah Ko · Soowon Son · +5 more
KAIST AI · Google DeepMind
Research Digest··2 min read
Hong and colleagues present FL4N, a model that turns monocular video into a representation for rendering a moving scene from different viewpoints and at different times.
The Daily Edition
The front page in your inbox each weekday morning: the lead, the picks, and why each one earned its place.