Data, training dataset and utilities - RRQuant: High-throughput quantification of seedling's epidermal integrity

This archive contains data and files related to the study : "RRQuant: High-throughput quantification of seedling's epidermal integrity" Traning and benchmarking dataset as well as model files for the hypocotyl segmentation models (rootpainter_compiled_datasets.rar) Data highlighting segmentation comparison between the different models (segmentation_comparison.zip) Code and utilities for model retraining from data proceesed by RRQuant (rootpainter_retraining_utilities.zip) Underlying data for figure 5, S1, S2 and S3 of the study (Figure5_data.zip, FigS1_data.zip and FigS2-3_data.zip) Content description for "rootpainter_compiled_datasets.rar". Root/├── Training/│ └── [V0 / V1 / V2]/ # Separate subdirectories for each model version│ ├── dataset/ # Raw images (split: training/validation)│ ├── masks/ # Ground truth masks (split: training/validation)│ └── model/ # Model file (.pkl)│└── Benchmarking/ ├── Benchmarking_dataset/ # Raw images for performance evaluation ├── Benchmarking_manual_masks/ # Ground-truth masks for the benchmark set └── Benchmarking_segmentation_output/ # Inference results from V0, V1, V2 for visual comparison Content description for "Figure5_data.zip" and "FigS1_data.zip": The folders rep1, rep2 and rep3 contain the original large tile scan merged images from the seedlings. The darkfield images (repX_dark) have been analyzed with the RRQuant workflow. The Brightfield images (repX_bf) facilitate the recognition of each condition position on the plate (sample names). The Split_Img folder contains images obtained during the RRQuant workflow and split based on the genotype, the staining and the replicate. It also contains the segmentation result for each split image and the quantification of morphological parameters and RR staining intensity (CSV files). Content description for "Figure5_data.zip" and "FigS1_data.zip": Contains data and prossesing (as for figure5 and S1) from samples corresponding to replicate 3 of Figure S1. The same samples have been images with 4 imaging systems: 1) A Leica M205CFA (same as the rest of the study, and same data as presented in rep3 of figure S1), 2) a Leica S9i, 3) a Leica MZ16 mounted with a 3D printed "OpenOcular" adaptor and a Samsung Galaxy A23 and 4) an EPSON Expression 12000XL image scanner.