ChronoRoot 2.0 MultiPlant
a model for plant root and shoot segmentation
1 min readupdated
Built on nnU-Net v2 (a biomedical image segmentation framework), this model is designed for multi-class separation in Arabidopsis, tomato, and Medicago images. The petiole is merged with the aerial class. The model was trained using SCNP (Same Class Neighbor Penalization to improve topological accuracy). It is well-suited for plant phenology and computer vision researchers working with plant image time series.
Key features
- Multi-class segmentation for Arabidopsis, tomatoes, and Medicago.
- Built on the nnU-Net v2 architecture.
- Trained using SCNP to enhance topological accuracy.
- The ready-to-use checkpoint is loaded from the
fold_0/checkpoint_final.pthfile.
What you can do
- Automatically segment plant roots and shoots in images.
- Analyze root and aerial system development in biological research.
- Integrate the model into image processing pipelines for plant time-lapse phenotyping.
Example
The model is loaded via the nnU-Net library to run image segmentation inference on plant images using a pre-trained checkpoint.
Read more
DetailsFramework: nnU-Net v2
Training method: SCNP (CVPR 2026)
Features: petiole merged with aerial class










Comments