, depth), and an autoregressive world model that predicts future scene states while separately modeling camera motion.
View-decoupled geometry states improve predictive 3D world modeling
By separating scene dynamics from camera-induced variation, GeoState outperforms prior models in depth forecasting and generalizes across driving datasets.
Chinese Tech
Hongchang Liu · Jian Liu · Yujuan Ding · Liang Peng · Xun Jiang · Fei Ma · +2 more
Tongji University · Hong Kong Polytechnic University · Tencent · Guangdong Laboratory of AI and Digital Economy (SZ)
Research Digest··3 min read
The authors propose GeoState, a view-decoupled geometry representation that disentangles scene content from camera viewpoint when using pretrained geometry foundation model (GFM) features as predictive states.
Why this paper
From Tencent and 3 others
In one line
GeoState separates scene state from camera viewpoint in geometry features, improving 3D world model prediction and cross-view querying.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§