Computer Vision Pipeline for Quantitative Analysis of Morphological Traits in Arabidopsis thaliana using RGB images

This repository provides the full source code used for the development of an RGB-based image pipeline for the extraction and analysis of morphological traits in A. thaliana at both the rosette and leaf levels.

Description

The RGB image analysis pipeline consists of several elements (see Figure 1): A – Data collection; B – Data preparation using auto-tray and auto-pot cropping; C – Annotation of whole plant canopies and individual leaves; D – Analysis of ecotype replicates based on growth and plant/soil colour distribution; E – Data splitting; F – Data augmentation based on affine transformations and colour jittering; G – Deep Learning-based rosette segmentation; H – Deep learning-based leaf segmentation and tracking; I – Example of basic geometrical plant traits used to calculate basic and derived numerical traits; J – Example of basic geometrical leaf traits used to calculate basic and derived numerical traits; K – Storage of extracted numerical plant- and leaf- level traits for each dataset; L – K-means clustering on the PCA-transformed datasets.

Figure 1. The complete workflow of the developed RGB image analysis pipeline for extracting and analysing various morphological traits of A. thaliana plants

Requirements

• python>=3.7
• torch>=1.4
• matplotlib
• numpy
• pandas
• pillow
• opencv-python
• scikit-learn
• jupyter notebook
• Microsoft Visual Studio >=2022

Structure

The fundamental filesystem structure resembles the tree shown below. Essentially, we have two main folders: code.

code
├───01_colour_n_area_analysis
│   ├───01_pre_processing
│   └───02_example_output
│
├───02_rosette_segmentation
│    ├───01_encoder_decoder_models_training.ipynb
│    ├───02_EDC_model_mask_generation.ipynb
│    ├───03_SAM_fine_tuning.ipynb
│    └───04_SAM_mask_generation.ipynb
│
├───03_leaf_segmentation_n_tracking
│   ├───01_train_instance_segmentation.ipynb
│   └───02_inference_tracking.ipynb
│   
├───04_GUI_based_plant_triats_extraction
│   ├───bin
│   │   └───Release
│   └───Properties
│
├───05_leaf_traits_extraction
│    ├───01_leaf_traits_calculations.ipynb
│    ├───02_calculated_leaf_tratis_plots.ipynb
│    └───03_leaf_excel_data.ipynb
│
└───06_clustering
    ├───01_plant_datastes_clust1.py
    └───02_leaf_datasets_clust1.py

How to use

Prerequisites

Each folder contains code for each task. Please go into each folder to find the required packages and install them before running the code. We recommend using Conda to manage the environments better.

conda create -n env_name 
conda activate env_name
pip install -f requirements.txt

LICENSE

This repo is distributed under LICENSE.