CellSeg3D: additional data & model weights

CellSeg3D additional benchmark data

Results files

Found in the `RESULTS_DATA` folder, the results of the benchmarking of the CellSeg3D pipeline on the test data are provided. The results are meant to be directly used by the ["Showing quantitative and qualitative benchmarking performance on additional datasets"](https://c-achard.github.io/cellseg3d-figures/Figure3/self-supervised-extra.html)

Simply point the notebook towards the `RESULTS_DATA` folder to use the results.

data_path = Path("path/to/RESULTS_DATA")

Additional model weights from re-training cellpose and StarDist models on the training data

Found in the `TRAINING` folder, the weights of the models used in the paper are provided. The training scripts are also provided for transparency.

Additional data for the CellSeg3D paper by Achard et al. Note : since Cellpose pre-trained performance was consistently higher than a retrained version, results were reported using the pre-trained model. See https://elifesciences.org/reviewed-preprints/99848v1 for more information.

Structure

RESULTS_DATA# do not move files, simply set the notebook path to use this folder

TRAINING # all training scripts and model weights

│

└───cellpose

│ │ cellpose_MoSk

│ │ | └───results # contains the weights of the model

│ │ | └───MouseSkull.py # script to train the model

│ │ cellpose_PlISH

│ │ | └───results # contains the weights of the model

│ │ | └───PlISH_CP.py # script to train the model

│ │ cellpose_PlNuc

│ │ | └───results # contains the weights of the model

│ │ | └───PlNuc_cellpose.py # script to train the model

└───stardist

│ MouseSKull

│ | └───models # contains the weights of the model

│ | └───MouseSkull_SD.py # script to train the model

│ PlISH

│ | └───models # contains the weights of the model

│ | └───PlatyISH_SD.py # script to train the model

│ PlNuc

│ | └───models # contains the weights of the model

│ | └───PlatyNuc_SD.py # script to train the model