Skip to content

Welcome to SpectralClusteringLink

ROMI_ICON2_greenB.png

For full documentation of the ROMI project visit docs.romi-project.eu.

AboutLink

A Python package designed to perform both semantic and instance segmentation of 3D plant point clouds, providing a robust and automatic pipeline for plant structure analysis.

OverviewLink

Spectral Clustering is a powerful tool for accurate segmentation and classification of plant 3D point clouds. By leveraging advanced graph-based methods, this package enables simultaneous semantic and instance segmentation, correcting potential segmentation defects and incorporating plant structural knowledge.

This method has been tested on both synthetic and real datasets to demonstrate reliability and efficiency.

Getting startedLink

To install the spectral_clustering conda package in an existing environment, first activate it, then proceed as follows: 

conda install spectral_clustering -c romi-eu

ContextLink

Accurate segmentation and classification of plants in 3D point clouds are essential for automated plant phenotyping. Traditional approaches rely on detecting plant organs based on local geometry but often overlook global plant structure.

Key FeaturesLink

  1. Point Scale Analysis

    • Utilizes similarity graphs to compute geometric attributes from the spectrum.
    • Distinguishes between linear organs (e.g., stems, branches, petioles) and planar organs (e.g., leaf blades).

  2. Organ Scale Analysis

    • Employs quotient graphs for detailed classification and to correct segmentation errors.
    • Maintains structural consistency of the plant.

ApplicationsLink

  • Synthetic and real 3D point cloud datasets of plants such as Chenopodium album (wild spinach) and Solanum lycopersicum (tomato plant).
  • Automatic pipelines for plant research.

BibliographyLink

This package is closely related to the following thesis:

Katia MIRANDE - Semantic and instance segmentation of plant 3D point cloud.
The work explores graph-based methods at point and organ scales for plant phenotyping. Full text can be accessed here.

Contribution & LicenseLink

We welcome contributions from the community! If you'd like to contribute, feel free to fork the repository or raise an issue.