# NVIDIA's Déjà View turns multi-angle photos into 3D geometry and camera positions

> Original title: Meet Déjà View, a model that turns photos taken from different angles into 3D geometry and camera positions.

- Source: NVIDIA AI
- Published: 2026-11-11T20:38:27.000Z
- Bad Influence: https://badinfluence.news/items/jm9fqvztsasizg0wyv1fek1ko
- Original: https://x.com/NVIDIAAI/status/2120617173286883692

## Summary

NVIDIA introduces Déjà View, a 117M-parameter model that reconstructs 3D geometry and camera positions from photos taken at different angles. By looping a single transformer block, the model matches or outperforms models up to 10 times larger across five benchmarks, using comparable or less compute. Researchers can adjust the number of loops to fit their compute budget, and the work is an oral paper at NeurIPS 2026 with code available.

## Post

Meet Déjà View, a model that turns photos taken from different angles into 3D geometry and camera positions.

By looping a single transformer block, this 117M-parameter model matches or outperforms models up to 10x larger across five benchmarks, using comparable or less compute. Researchers can dial the number of loops to fit their compute budget.

#NeurIPS2026 oral paper + code: http://research.nvidia.com/labs/dvl/projects/dvlt

[Video](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Famplify_video_thumb%2F2120616346136563712%2Fimg%2FdOYqfqajmpOffWHJ.jpg&mode=thumb&exp=1794700800&sig=0c5367dc2a7b21e0)