ngaggion's picture
Add the SCNP-trained MultiPlant model
cee42e4 verified
|
Raw History Blame Contribute Delete
2.32 kB
metadata
license: mit
library_name: nnunet
tags:
  - image-segmentation
  - biology
  - chronoroot
  - plant-science
datasets:
  - ngaggion/ChronoRoot2

ChronoRoot 2.0: MultiPlant Segmentation Model

Pre-trained nnU-Net v2 model for multi-class segmentation of Arabidopsis, tomato, and Medicago images. ChronoRoot loads fold_0/checkpoint_final.pth. Petiole is merged into the aerial class.

The model was trained with SCNP (Same Class Neighbor Penalization; Valverde, Papadopoulos, Larsen, and Dahl, CVPR 2026). SCNP is used only during training. The published checkpoint is a standard nnU-Net model. Some Medicago images were included in this training and will be made public shortly.

The Arabidopsis-only model is ChronoRoot2-Arabidopsis.

The same videos as the annotated set, sampled every hour and without annotations, are at UnsupervisedRootVideos.

ChronoRoot: website and GitHub.

Please cite ChronoRoot 2.0 if you use this model:

@article{gaggion2026chronoroot,
  title={ChronoRoot 2.0: An Open AI-Powered Platform for 2D Temporal Plant Phenotyping},
  author={Nicolás Gaggion and Noelia A. Boccardo and Rodrigo Bonazzola and María Florencia Legascue and María Florencia Mammarella and Florencia Sol Rodriguez and Federico Emanuel Aballay and Florencia Belén Catulo and Andana Barrios and Luciano J. Santoro and Franco Accavallo and Santiago Nahuel Villarreal and Leonardo I. Pereyra-Bistrain and Moussa Benhamed and Martin Crespi and Martiniano María Ricardi and Ezequiel Petrillo and Thomas Blein and Federico Ariel and Enzo Ferrante},
  journal={GigaScience},
  volume={15},
  pages={giag018},
  year={2026},
  doi={10.1093/gigascience/giag018}
}

SCNP:

@inproceedings{valverde2026scnp,
  title={Towards High-Quality Image Segmentation: Improving Topology Accuracy by Penalizing Neighbor Pixels},
  author={Valverde, Juan Miguel and Papadopoulos, Dim P. and Larsen, Rasmus and Dahl, Anders},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2026},
  url={https://arxiv.org/abs/2603.18671}
}