Geometry Decoder
A component within GeoState that queries scene state using camera rays to render observations from a specific viewpoint. The camera serves as a query condition during decoding rather than as a constituent part of the state, allowing the decoder to produce view-specific renderings from the underlying scene representation.
Timeline (last 90 days)
GeoState contains three key components: a scene tokenizer, a geometry decoder and an autoregressive world model; the scene tokenizer aggregates multi-view GFM features into a viewpoint-decoupled scene state; the geometry decoder queries the scene state with camera rays to render observations from a specific viewpoint; the autoregressive world model performs prediction in the decoupled scene state space and models camera motion independently.
The geometry decoder uses camera rays as queries, sampling information from the scene state in a manner similar to ray casting or cross-attention, to generate geometric observations (such as depth maps, feature maps or RGB images) from the target camera viewpoint.
All Facts (2)
GeoState contains three key components: a scene tokenizer, a geometry decoder and an autoregressive world model; the scene tokenizer aggregates multi-view GFM features into a viewpoint-decoupled scene state; the geometry decoder queries the scene state with camera rays to render observations from a specific viewpoint; the autoregressive world model performs prediction in the decoupled scene state space and models camera motion independently.
50%UnverifiedThe geometry decoder uses camera rays as queries, sampling information from the scene state in a manner similar to ray casting or cross-attention, to generate geometric observations (such as depth maps, feature maps or RGB images) from the target camera viewpoint.
50%