English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Zero-shot novel view synthesis (NVS) from a single image is an essentialproblem in 3D object understanding. While recent approaches that leveragepre-trained generative models can synthesize high-quality novel views fromin-the-wild inputs, they still struggle to maintain 3D consistency acrossdifferent views. In this paper, we present Consistent-1-to-3, which is agenerative framework that significantly mitigate this issue. Specifically, wedecompose the NVS task into two stages: (i) transforming observed regions to anovel view, and (ii) hallucinating unseen regions. We design a scenerepresentation transformer and view-conditioned diffusion model for performingthese two stages respectively. Inside the models, to enforce 3D consistency, wepropose to employ epipolor-guided attention to incorporate geometryconstraints, and multi-view attention to better aggregate multi-viewinformation. Finally, we design a hierarchy generation paradigm to generatelong sequences of consistent views, allowing a full 360 observation of theprovided object image. Qualitative and quantitative evaluation over multipledatasets demonstrate the effectiveness of the proposed mechanisms againststate-of-the-art approaches. Our project page is athttps://jianglongye.com/consistent123/

Consistent-1-to-3: Consistent Image to 3D View Synthesis via Geometry-aware Diffusion Models