StyleTRF: Stylizing Tensorial Radiance Fields

TL;DR

StyleTRF introduces a fast, efficient method for stylized view generation using TensoRF, decoupling style adaptation from scene capture and outperforming previous methods in speed and quality.

Contribution

It proposes a novel, quick-to-optimize approach for scene stylization with TensoRF, reducing training time and simplifying the process compared to prior joint optimization methods.

Findings

Achieves state-of-the-art stylization results on multiple scenes.

Significantly reduces training time compared to SNeRF-based methods.

Effectively decouples style adaptation from scene geometry capture.

Abstract

Stylized view generation of scenes captured casually using a camera has received much attention recently. The geometry and appearance of the scene are typically captured as neural point sets or neural radiance fields in the previous work. An image stylization method is used to stylize the captured appearance by training its network jointly or iteratively with the structure capture network. The state-of-the-art SNeRF method trains the NeRF and stylization network in an alternating manner. These methods have high training time and require joint optimization. In this work, we present StyleTRF, a compact, quick-to-optimize strategy for stylized view generation using TensoRF. The appearance part is fine-tuned using sparse stylized priors of a few views rendered using the TensoRF representation for a few iterations. Our method thus effectively decouples style-adaption from view capture and is much faster than the previous methods. We show state-of-the-art results on several scenes used for this purpose.