GrowSplat: Constructing Temporal Digital Twins of Plants with Gaussian Splats

TL;DR

GrowSplat introduces a new framework combining 3D Gaussian Splatting and a two-stage registration pipeline to create accurate 4D models of plant growth from multi-view data, aiding plant phenotyping.

Contribution

The paper presents a novel method for constructing temporal digital twins of plants using Gaussian Splatting and robust registration, addressing challenges of complex geometries and deformations.

Findings

Effective reconstruction of plant growth over time.

Detailed 4D models of Sequoia and Quinoa species.

Robust alignment pipeline improves temporal consistency.

Abstract

Accurate temporal reconstructions of plant growth are essential for plant phenotyping and breeding, yet remain challenging due to complex geometries, occlusions, and non-rigid deformations of plants. We present a novel framework for building temporal digital twins of plants by combining 3D Gaussian Splatting with a robust sample alignment pipeline. Our method begins by reconstructing Gaussian Splats from multi-view camera data, then leverages a two-stage registration approach: coarse alignment through feature-based matching and Fast Global Registration, followed by fine alignment with Iterative Closest Point. This pipeline yields a consistent 4D model of plant development in discrete time steps. We evaluate the approach on data from the Netherlands Plant Eco-phenotyping Center, demonstrating detailed temporal reconstructions of Sequoia and Quinoa species. Videos and Images can be seen at https://berkeleyautomation.github.io/GrowSplat/