BADGR: Bundle Adjustment Diffusion Conditioned by GRadients for Wide-Baseline Floor Plan Reconstruction

Alex Colburn, Yuguang Li, Ivaylo Boyadzhiev, Zixuan Liu, Linda Shapiro

Abstract:

Reconstructing precise camera poses and floor plan lay-
outs from wide-baseline RGB panoramas is a difficult and
unsolved problem. We introduce BADGR, a novel diffu-
sion model that jointly performs reconstruction and bundle
adjustment (BA) to refine poses and layouts from a coarse
state, using 1D floor boundary predictions from dozens of
sparsely captured images. Unlike guided diffusion models,
BADGR is conditioned on dense per-column outputs from
a single-step Levenberg Marquardt (LM) optimizer and is
trained to predict camera and wall positions, while minimiz-
ing reprojection errors for view consistency. The objective
of layout generation from denoising diffusion process com-
plements BA optimization by providing additional learned
layout-structural constraints on top of the co-visible fea-
tures across images. These constraints help BADGR make
plausible guesses about spatial relationships, which con-
strain the pose graph, such as wall adjacency and collinear-
ity, while also learning to mitigate errors from dense bound-
ary observations using global context. BADGR trains exclu-
sively on 2D floor plans, simplifying data acquisition, en-
abling robust augmentation, and supporting a variety of in-
put densities. Our experiments validate our method, which
significantly outperforms the state-of-the-art pose and floor
plan layout reconstruction with different input densities.