Self-improving Multiplane-to-layer Images for Novel View Synthesis

TL;DR

This paper introduces a lightweight, fast, and scene-agnostic method for novel view synthesis that converts fronto-parallel planes into deformable layers and refines them with a feed-forward process, outperforming recent models.

Contribution

It proposes a novel multiplane-to-layer conversion and refinement approach that eliminates the need for scene-specific tuning and improves inference speed and memory efficiency.

Findings

Outperforms recent models in metrics and human evaluation.

Does not require fine-tuning for new scenes.

Offers faster inference and more compact layered representations.

Abstract

We present a new method for lightweight novel-view synthesis that generalizes to an arbitrary forward-facing scene. Recent approaches are computationally expensive, require per-scene optimization, or produce a memory-expensive representation. We start by representing the scene with a set of fronto-parallel semitransparent planes and afterward convert them to deformable layers in an end-to-end manner. Additionally, we employ a feed-forward refinement procedure that corrects the estimated representation by aggregating information from input views. Our method does not require fine-tuning when a new scene is processed and can handle an arbitrary number of views without restrictions. Experimental results show that our approach surpasses recent models in terms of common metrics and human evaluation, with the noticeable advantage in inference speed and compactness of the inferred layered geometry, see https://samsunglabs.github.io/MLI