We propose a Pose-Free Large Reconstruction Model (PF-LRM) for reconstructing a 3D
object from a few unposed images even with little visual overlap, while simultaneously …

Abstract Structure-from-motion (SfM) is a long-standing problem in the computer vision
community which aims to reconstruct the camera poses and 3D structure of a scene from a …

The image matching field has been witnessing a continuous emergence of novel learnable
feature matching techniques with ever-improving performance on conventional benchmarks …

We address the task of estimating 6D camera poses from sparse-view image sets (2-8
images). This task is a vital pre-processing stage for nearly all contemporary (neural) …

Open-world 3D generation has recently attracted considerable attention. While many single-
image-to-3D methods have yielded visually appealing outcomes, they often lack sufficient …

Are camera poses necessary for multi-view 3D modeling? Existing approaches
predominantly assume access to accurate camera poses. While this assumption might hold …

Despite the significant progress in six degrees-of-freedom (6DoF) object pose estimation
existing methods have limited applicability in real-world scenarios involving embodied …

Estimating relative camera poses between images has been a central problem in computer
vision. Methods that find correspondences and solve for the fundamental matrix offer high …

Implicit neural representations for video (NeRV) have recently become a novel way for high-
quality video representation. However existing works employ a single network to represent …

The practicality of 3D object pose estimation remains limited for many applications due to
the need for prior knowledge of a 3D model and a training period for new objects. To …