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

While object reconstruction has made great strides in recent years, currentmethods typically require densely captured images and/or known camera poses,and generalize poorly to novel object categories. To step toward objectreconstruction in the wild, this work explores reconstructing generalreal-world objects from a few images without known camera poses or objectcategories. The crux of our work is solving two fundamental 3D vision problems-- shape reconstruction and pose estimation -- in a unified approach. Ourapproach captures the synergies of these two problems: reliable camera poseestimation gives rise to accurate shape reconstruction, and the accuratereconstruction, in turn, induces robust correspondence between different viewsand facilitates pose estimation. Our method FORGE predicts 3D features fromeach view and leverages them in conjunction with the input images to establishcross-view correspondence for estimating relative camera poses. The 3D featuresare then transformed by the estimated poses into a shared space and are fusedinto a neural radiance field. The reconstruction results are rendered by volumerendering techniques, enabling us to train the model without 3D shapeground-truth. Our experiments show that FORGE reliably reconstructs objectsfrom five views. Our pose estimation method outperforms existing ones by alarge margin. The reconstruction results under predicted poses are comparableto the ones using ground-truth poses. The performance on novel testingcategories matches the results on categories seen during training. Projectpage: https://ut-austin-rpl.github.io/FORGE/

Few-View Object Reconstruction with Unknown Categories and Camera Poses