Pretrained nucleAIzer models for microscopy datasets

This upload consists of 4 models trained using the nucleAIzer algorithm (Hollandi et al. & https://github.com/spreka/biomagdsb) for segmenting the following microscopy image types (profiles):

1. general nucleus
2. fluorescent nucleus
3. tiny&clumped fluorescent nucleus
4. tissue nucleus
5. fluorescent cytoplasm
6. tissue nucleus IHC
7. tissue nucleus H&E

Models for each type:

1. mask_rcnn_general.h5
2. mask_rcnn_general.h5
3. mask_rcnn_general.h5
4. mask_rcnn_general.h5
5. mask_rcnn_cytoplasm.h5
6. mask_rcnn_tissue_bluebrown.h5
7. mask_rcnn_tissue_pink.h5

The above models were trained on resized images such that  the median nuclei diameter is 40 pixels.

Additionally, a presegmentation model is also attached (mask_rcnn_presegmentation.h5) for generating the weak labels on the test set (first step in the training pipeline), while the mask_rcnn_fluo_simple.h5 is a general model that can be used for automatic object size estimation.

We additionally attach a profiles.json file describing the parameters for the models for the above segmentation tasks that can be loaded by our prediction and training code used in the napari nucleAIzer plugin (https://pypi.org/project/napari-nucleaizer/ and https://pypi.org/project/nucleaizer-backend/).

If you use our models in your research work, please cite our original paper (Hollandi et al.: nucleAIzer: a parameter-free deep learning framework for nucleus segmentation using image style transfer, Cell Systems, 2020. https://doi.org/10.1016/j.cels.2020.04.003)