Totem first instantiates each selected object from a reference image as a 3D memory, comprising a canonical representation with surface anchors and an object-to-world transform that places it in a shared coordinate system.
Object-centric 3D memory enables consistent video generation under large viewpoint changes
Totem maintains persistent 3D object states and uses dual-stream conditioning with object-centric attention to stabilize appearance, geometry, and spatial relations across generated views.
Chinese Tech
Zichen Wan · Tianyu Huang · Fan Li · Kai Wang · Hongzhi Zhang · Wangmeng Zuo
Harbin Institute of Technology · Huawei Technologies Ltd.
Research Digest··2 min read
The authors introduce Totem, a framework that represents each selected object as a persistent 3D memory in a shared world frame, combining object memories with scene geometry for target-view synthesis.
Why this paper
From Huawei Technologies Ltd. and Harbin Institute of Technology
In one line
Totem uses object-centric 3D memories to generate videos with consistent objects across large viewpoint changes.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·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.
§