Reproduce “DualPM: Dual Posed-Canonical Point Maps for 3D Shape and Pose Reconstruction” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Ben Kaye, Tomas Jakab, Shangzhe Wu, Christian Rupprecht, Andrea Vedaldi
DualPM reconstructs deformable quadrupeds from a single image by predicting, for each pixel, 3D points in the observed and canonical poses. Their paired maps encode deformation and correspondence, while layered predictions address self-occlusion. On horse keypoint transfer, it reaches 73.2% [email protected] versus 53.9% for 3D-Fauna; results are limited by mask input, synthetic training, and ambiguous unseen regions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.