RootQuantV2: Adapting a Vision Foundation Model for Root-Trait Regression from Minirhizotron Imagery
Abstract
A lack of high-throughput phenotyping solutions for root traits in field-grown crops has severely constrained understanding and improvement of below-ground traits and processes. Minirhizotrons are the standard non-destructive root-phenotyping method in field environments. Computer vision solutions are needed to allow automated trait estimation at scale, but training data is scarce and human annotations are often inaccessible because they reside in proprietary software that only exports per-image scalar totals of root length and surface area. Nevertheless, large numeric archives of these root traits already exist. RootQuant showed that the traits can be predicted directly from the whole image by regression, thus removing manually traced masks from the pipeline; RootQuantV2 takes that idea further by replacing RootQuant's CNN backbone with a self-supervised ViT. We adapt a frozen DINOv3 ViT-L/16 with a hybrid parameter-efficient scheme. Training only 11.9M parameters (3.78% of the model), RootQuantV2 achieves length and area R2 of 0.950 and 0.930, respectively, while lowering length/area RMSE by 24.3%/20.7% over RootQuant. RootQuantV2 thus repurposes legacy numeric archives for high-throughput, automated root trait estimation.
Related
- Matched-Input Estimates Differ in Sign Across Architectures: Auditing EEG Foundation Models on Motor Imagery
- Adapting Boltz-2 with limited experimental activity data improves early enrichment in virtual screening
- Frequency-dependent fitness effects are ubiquitous in rapidly adapting microbial populations
- Pre-season bare-soil spectral proxies complement pod-filling canopy indices for field-scale soybean-yield prediction from UAV imagery
- Multi-scale vision transformer model with UAV-based hyperspectral imagery for early identification of pine wilt disease infected trees
BibTeX
@misc{parth2026rootquantv2,
title = {{RootQuantV2: Adapting a Vision Foundation Model for Root-Trait Regression from Minirhizotron Imagery}},
author = {Kinjalk Parth and Sebastian Varela and Andrew D. B. Leakey},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.25567},
eprint = {2609.25567},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.25567},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - RootQuantV2: Adapting a Vision Foundation Model for Root-Trait Regression from Minirhizotron Imagery
AU - Kinjalk Parth
AU - Sebastian Varela
AU - Andrew D. B. Leakey
PY - 2026
DA - 2026/09/22
PB - arXiv
DO - 10.48550/arxiv.2609.25567
UR - https://arxiv.org/abs/2609.25567
AB - A lack of high-throughput phenotyping solutions for root traits in field-grown crops has severely constrained understanding and improvement of below-ground traits and processes. Minirhizotrons are the standard non-destructive root-phenotyping method in field environments. Computer vision solutions are needed to allow automated trait estimation at scale, but training data is scarce and human annotations are often inaccessible because they reside in proprietary software that only exports per-image scalar totals of root length and surface area. Nevertheless, large numeric archives of these root traits already exist. RootQuant showed that the traits can be predicted directly from the whole image by regression, thus removing manually traced masks from the pipeline; RootQuantV2 takes that idea further by replacing RootQuant's CNN backbone with a self-supervised ViT. We adapt a frozen DINOv3 ViT-L/16 with a hybrid parameter-efficient scheme. Training only 11.9M parameters (3.78% of the model), RootQuantV2 achieves length and area R^2 of 0.950 and 0.930, respectively, while lowering length/area RMSE by 24.3%/20.7% over RootQuant. RootQuantV2 thus repurposes legacy numeric archives for high-throughput, automated root trait estimation.
ER -