SceneRF: Self-Supervised Monocular 3D Scene Reconstruction with Radiance Fields

TL;DR

SceneRF is a self-supervised monocular 3D scene reconstruction method that leverages neural radiance fields and explicit depth optimization, eliminating the need for depth supervision during training.

Contribution

It introduces a novel self-supervised approach using only posed image sequences and a probabilistic sampling strategy for efficient large scene reconstruction.

Findings

Outperforms baselines in novel depth view synthesis

Achieves superior scene reconstruction on indoor and outdoor datasets

Operates with only monocular images without depth supervision

Abstract

3D reconstruction from a single 2D image was extensively covered in the literature but relies on depth supervision at training time, which limits its applicability. To relax the dependence to depth we propose SceneRF, a self-supervised monocular scene reconstruction method using only posed image sequences for training. Fueled by the recent progress in neural radiance fields (NeRF) we optimize a radiance field though with explicit depth optimization and a novel probabilistic sampling strategy to efficiently handle large scenes. At inference, a single input image suffices to hallucinate novel depth views which are fused together to obtain 3D scene reconstruction. Thorough experiments demonstrate that we outperform all baselines for novel depth views synthesis and scene reconstruction, on indoor BundleFusion and outdoor SemanticKITTI. Code is available at https://astra-vision.github.io/SceneRF .