Bolt3D: Generating 3D Scenes in Seconds
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866918162161729536 |
|---|---|
| author | Szymanowicz, Stanislaw Zhang, Jason Y. Srinivasan, Pratul Gao, Ruiqi Brussee, Arthur Holynski, Aleksander Martin-Brualla, Ricardo Barron, Jonathan T. Henzler, Philipp |
| author_facet | Szymanowicz, Stanislaw Zhang, Jason Y. Srinivasan, Pratul Gao, Ruiqi Brussee, Arthur Holynski, Aleksander Martin-Brualla, Ricardo Barron, Jonathan T. Henzler, Philipp |
| contents | We present a latent diffusion model for fast feed-forward 3D scene generation. Given one or more images, our model Bolt3D directly samples a 3D scene representation in less than seven seconds on a single GPU. We achieve this by leveraging powerful and scalable existing 2D diffusion network architectures to produce consistent high-fidelity 3D scene representations. To train this model, we create a large-scale multiview-consistent dataset of 3D geometry and appearance by applying state-of-the-art dense 3D reconstruction techniques to existing multiview image datasets. Compared to prior multiview generative models that require per-scene optimization for 3D reconstruction, Bolt3D reduces the inference cost by a factor of up to 300 times. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_14445 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Bolt3D: Generating 3D Scenes in Seconds Szymanowicz, Stanislaw Zhang, Jason Y. Srinivasan, Pratul Gao, Ruiqi Brussee, Arthur Holynski, Aleksander Martin-Brualla, Ricardo Barron, Jonathan T. Henzler, Philipp Computer Vision and Pattern Recognition We present a latent diffusion model for fast feed-forward 3D scene generation. Given one or more images, our model Bolt3D directly samples a 3D scene representation in less than seven seconds on a single GPU. We achieve this by leveraging powerful and scalable existing 2D diffusion network architectures to produce consistent high-fidelity 3D scene representations. To train this model, we create a large-scale multiview-consistent dataset of 3D geometry and appearance by applying state-of-the-art dense 3D reconstruction techniques to existing multiview image datasets. Compared to prior multiview generative models that require per-scene optimization for 3D reconstruction, Bolt3D reduces the inference cost by a factor of up to 300 times. |
| title | Bolt3D: Generating 3D Scenes in Seconds |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2503.14445 |