Pre-trained Stardist 3D models from paper '3D Nuclei Segmentation by Combining GAN Based Image Synthesis and Existing 3D Manual Annotations'
Authors/Creators
-
1.
Institut Interdisciplinaire de Neuroscience
-
2.
Bordeaux Imaging Center
Description
This contains the 3 Stardist 3D models trained for the paper '3D Nuclei Segmentation by Combining GAN Based Image Synthesis and Existing 3D Manual Annotations'.
model_sospim.zip contains the weights of the model trained with the original sospim fluorescent images (stained with DAPI and SOX2 labels) and the manual annotations. The average size of one nucleus in pixel is [x=27, y=28, z=10].
model_confocal.zip contains the weights of the model trained with the synthethizes modules mimicking the confocal acquisitions (stained with the FUCCI label) and the manual annotations. The average size of one nucleus in pixel is [x=39, y=39, z=7].
model_spinning.zip contains the weights of the model trained with the synthethizes modules mimicking the spinning-disk acquisitions (stained with the DAPI label) and the manual annotations. The average size of one nucleus in pixel is [x=39, y=39, z=7].