Diffusion-Based Synthetic Pretraining for Root Segmentation in Rhizotron Images
- 1Wageningen University and Research
Abstract
Rhizotrons allow root system architecture to be studied non-destructively in soil. However, automated analysis of rhizotron images remains difficult: roots have low contrast against soil, soil particles resemble fine roots, and deep learning models need many manually annotated images that are slow and subjective to produce. Synthetic images could provide paired images and annotations at scale, but rhizotron images are largely absent from the data used to train generative models. We built a pipeline combining Low-Rank Adaptation (LoRA), ControlNet and inpainting to generate synthetic rhizotron images with paired root masks. We compared training on synthetic data only, real data only, and synthetic pretraining followed by fine-tuning on real images, using three segmentation models: nnU-Net v2, YOLO26 and Mask2Former. Synthetic-only training reached a Dice score of up to 0.60. Synthetic pretraining followed by fine-tuning on the full real training set (two annotated rhizotron images) reached 0.80, and nnU-Net v2 and Mask2Former reached 0.62 with only two annotated image patches containing visible roots. When real annotations were limited, pretraining also improved segmentation of unseen crops, including lettuce, barley, wheat, maize, tomato, cucumber, French bean and field bean. We also introduced a Soft Dice score that compares predictions with probabilistic consensus maps from seven independent annotators, accounting for annotator disagreement. Models matched average human agreement on low-ambiguity patches but remained below it on high-ambiguity patches. In conclusion, diffusion-generated images can complement real annotations and help scale root segmentation, although their effect on downstream root trait measurements still needs to be validated.
Related
- Texture-aware diffusion modelling for severity-conditioned synthetic progression-like visualization of diabetic foot ulcers from single images
- Deep Learning Based Cell Segmentation of 3D Confocal Images of Wood for Enzymatic Hydrolysis Applications
- Automatic Generation of Expert-Level Neuron Segmentation Masks from Fluorescence Microscopy Images for Non-Invasive Deep Learning Analysis of Phase-Contrast Images
- ACMNet: An attention-convolution mixed neural network for semantic segmentation of high-resolution remote sensing images
- Synthetic-augmented multimodal deep learning fuses dual-angle RGB images and phenology to unlock genotype-informative canopy structural trait in wheat
BibTeX
@misc{lu2026diffusionbased,
title = {{Diffusion-Based Synthetic Pretraining for Root Segmentation in Rhizotron Images}},
author = {F. Lu and G. Kootstra and Z. K. J. Hartley and R. van de Zedde and M. P. Pound and V. Willemsen},
howpublished = {bioRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.30.763050},
url = {https://www.biorxiv.org/content/10.64898/2026.10.30.763050v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Diffusion-Based Synthetic Pretraining for Root Segmentation in Rhizotron Images
AU - F. Lu
AU - G. Kootstra
AU - Z. K. J. Hartley
AU - R. van de Zedde
AU - M. P. Pound
AU - V. Willemsen
PY - 2026
DA - 2026/10/30
PB - bioRxiv
DO - 10.64898/2026.10.30.763050
UR - https://www.biorxiv.org/content/10.64898/2026.10.30.763050v1
AB - Rhizotrons allow root system architecture to be studied non-destructively in soil. However, automated analysis of rhizotron images remains difficult: roots have low contrast against soil, soil particles resemble fine roots, and deep learning models need many manually annotated images that are slow and subjective to produce. Synthetic images could provide paired images and annotations at scale, but rhizotron images are largely absent from the data used to train generative models. We built a pipeline combining Low-Rank Adaptation (LoRA), ControlNet and inpainting to generate synthetic rhizotron images with paired root masks. We compared training on synthetic data only, real data only, and synthetic pretraining followed by fine-tuning on real images, using three segmentation models: nnU-Net v2, YOLO26 and Mask2Former. Synthetic-only training reached a Dice score of up to 0.60. Synthetic pretraining followed by fine-tuning on the full real training set (two annotated rhizotron images) reached 0.80, and nnU-Net v2 and Mask2Former reached 0.62 with only two annotated image patches containing visible roots. When real annotations were limited, pretraining also improved segmentation of unseen crops, including lettuce, barley, wheat, maize, tomato, cucumber, French bean and field bean. We also introduced a Soft Dice score that compares predictions with probabilistic consensus maps from seven independent annotators, accounting for annotator disagreement. Models matched average human agreement on low-ambiguity patches but remained below it on high-ambiguity patches. In conclusion, diffusion-generated images can complement real annotations and help scale root segmentation, although their effect on downstream root trait measurements still needs to be validated.
ER -