SPROUT: A Scalable Diffusion Foundation Model for Agricultural Vision

TL;DR

SPROUT is a scalable diffusion-based foundation model specifically designed for agricultural vision tasks, trained on a large dataset, outperforming existing models with lower pre-training costs.

Contribution

Introduces SPROUT, a novel diffusion transformer model for agriculture, trained on 2.6 million images, achieving superior performance over prior models.

Findings

SPROUT outperforms state-of-the-art models on various agricultural tasks.

It requires significantly less pre-training cost than existing models.

The model effectively learns structure-aware representations through diffusion denoising.

Abstract

Vision Foundation Models (VFM) pre-trained on large-scale unlabeled data have achieved remarkable success on general computer vision tasks, yet typically suffer from significant domain gaps when applied to agriculture. In this context, we introduce $SPROUT$ ($S$calable $P$lant $R$epresentation model via $O$pen-field $U$nsupervised $T$raining), a multi-crop, multi-task agricultural foundation model trained via diffusion denoising. SPROUT leverages a VAE-free Pixel-space Diffusion Transformer to learn rich, structure-aware representations through denoising and enabling efficient end-to-end training. We pre-train SPROUT on a curated dataset of 2.6 million high-quality agricultural images spanning diverse crops, growth stages, and environments. Extensive experiments demonstrate that SPROUT consistently outperforms state-of-the-art web-pretrained and agricultural foundation models across a wide range of downstream tasks, while requiring substantially lower pre-training cost. The code and model are available at https://github.com/UTokyo-FieldPhenomics-Lab/SPROUT.