MetaView: Monocular Novel View Synthesis with Scale-Aware Implicit Geometry Priors
MetaView aims to render large viewpoint changes from one image by mixing implicit geometry with metric depth.
The paper argues that explicit geometry helps consistency but can limit generalization, while implicit scene modeling is more flexible but weaker on camera control. MetaView uses a feed-forward geometry perception network for implicit geometry priors, then anchors generation with metric depth. The authors say experiments show stronger geometry consistency, controllability, and generalization under challenging monocular view changes. Code is listed as publicly available. HF Daily Papers' note
The paper argues that explicit geometry helps consistency but can limit generalization, while implicit scene modeling is more flexible but weaker on camera control. MetaView uses a feed-forward geometry perception network for implicit geometry priors, then anchors generation with metric depth. The authors say experiments show stronger geometry consistency, controllability, and generalization under challenging monocular view changes. Code is listed as publicly available. HF Daily Papers' note
score 4