Improve 3D object reconstruction using unsupervised learning on image collection with TARS
Improve 3D object reconstruction using unsupervised learning on image collection with TARS
Topologically-Aware Deformation Fields for Single-View 3D Reconstruction
arXiv paper abstract https://arxiv.org/abs/2205.06267
arXiv PDF paper https://shivamduggal4.github.io/tars-3D/static/paper/main.pdf
Supplementary PDF paper https://shivamduggal4.github.io/tars-3D/static/paper/supplementary.pdf
Project page https://shivamduggal4.github.io/tars-3D
YouTube https://www.youtube.com/watch?v=hSWaN_6Ir1c
… present a new framework for learning 3D object shapes and dense cross-object 3D correspondences from just an unaligned category-specific image collection.
The 3D shapes are generated … as deformations to a category-specific signed distance field and are learned in an unsupervised manner solely from unaligned image collections without any 3D supervision.
… image collections … contain several intra-category … variations … Because of this, prior works either focus on … shape individually without modeling … correspondences or perform … correspondence … on categories with minimal … variations.
… overcome these restrictions by learning a topologically-aware implicit deformation field that maps a 3D point in the object space to a higher dimensional point in the category-specific canonical space.
At inference … given a single image, … reconstruct the underlying 3D shape by first implicitly deforming each 3D point in the object space to the learned category-specific canonical space using the topologically-aware deformation field and then reconstructing the 3D shape as a canonical signed distance field.
… approach … TARS, achieves state-of-the-art reconstruction fidelity on several datasets: ShapeNet, Pascal3D+, CUB, and Pix3D chairs. …
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
