Get 3D object shape and new view from one image by geometry constraints with Consistent-1-to-3
Get 3D object shape and new view from one image by geometry constraints with Consistent-1-to-3
Consistent-1-to-3: Consistent Image to 3D View Synthesis via Geometry-aware Diffusion Models
arXiv paper abstract https://arxiv.org/abs/2310.03020
arXiv PDF paper https://arxiv.org/pdf/2310.03020.pdf
Project page https://jianglongye.com/consistent123
Zero-shot novel view synthesis (NVS) from a single image is an essential problem in 3D object understanding … While … generative models can synthesize … novel views … they still struggle to maintain 3D consistency across different views.
… present Consistent-1-to-3, which is a generative framework … mitigate this issue … decompose the NVS task into two stages: (i) transforming observed regions to a novel view, and (ii) hallucinating unseen regions.
… design a scene representation transformer and view-conditioned diffusion model for performing these two stages respectively.
… to enforce 3D consistency, … propose to employ epipolor-guided attention to incorporate geometry constraints, and multi-view attention to better aggregate multi-view information.
… design a hierarchy generation paradigm to generate long sequences of consistent views, allowing a full 360 observation of the provided object image.
… demonstrate the effectiveness of the proposed mechanisms against state-of-the-art approaches …
Stay up to date. Subscribe to my posts https://morrislee1234.wixsite.com/website/contact
Web site with my other posts by category https://morrislee1234.wixsite.com/website
LinkedIn https://www.linkedin.com/in/morris-lee-47877b7b
