RGBD-Net: Predicting color and depth images for novel views synthesis

TL;DR

RGBD-Net is a novel cascaded architecture for synthesizing high-quality color and depth images for new viewpoints, outperforming existing methods and generalizing well without scene-specific tuning.

Contribution

The paper introduces RGBD-Net, combining hierarchical depth regression with a depth-aware generator, enabling superior view synthesis and dense 3D point cloud creation without per-scene optimization.

Findings

Outperforms state-of-the-art in view synthesis

Generalizes well to new scenes without per-scene optimization

Can be trained without depth supervision while maintaining quality

Abstract

We propose a new cascaded architecture for novel view synthesis, called RGBD-Net, which consists of two core components: a hierarchical depth regression network and a depth-aware generator network. The former one predicts depth maps of the target views by using adaptive depth scaling, while the latter one leverages the predicted depths and renders spatially and temporally consistent target images. In the experimental evaluation on standard datasets, RGBD-Net not only outperforms the state-of-the-art by a clear margin, but it also generalizes well to new scenes without per-scene optimization. Moreover, we show that RGBD-Net can be optionally trained without depth supervision while still retaining high-quality rendering. Thanks to the depth regression network, RGBD-Net can be also used for creating dense 3D point clouds that are more accurate than those produced by some state-of-the-art multi-view stereo methods.