Scene Coordinate Reconstruction: Posing of Image Collections via Incremental Learning of a Relocalizer
This addresses camera pose estimation for 3D reconstruction, offering a learning-based alternative that does not require pose priors or sequential inputs, though it is incremental in nature.
The paper tackles camera pose estimation from unposed images by reinterpreting incremental structure-from-motion as an iterative refinement of a learning-based relocalizer, achieving accuracy close to feature-based methods in many cases, as shown through novel view synthesis.
We address the task of estimating camera parameters from a set of images depicting a scene. Popular feature-based structure-from-motion (SfM) tools solve this task by incremental reconstruction: they repeat triangulation of sparse 3D points and registration of more camera views to the sparse point cloud. We re-interpret incremental structure-from-motion as an iterated application and refinement of a visual relocalizer, that is, of a method that registers new views to the current state of the reconstruction. This perspective allows us to investigate alternative visual relocalizers that are not rooted in local feature matching. We show that scene coordinate regression, a learning-based relocalization approach, allows us to build implicit, neural scene representations from unposed images. Different from other learning-based reconstruction methods, we do not require pose priors nor sequential inputs, and we optimize efficiently over thousands of images. In many cases, our method, ACE0, estimates camera poses with an accuracy close to feature-based SfM, as demonstrated by novel view synthesis. Project page: https://nianticlabs.github.io/acezero/