3D Mask R-CNN data

Image repository of the paper 3D Mask R-CNN Benchmarks in Controlled Environment and Morphogenesis Study
This repository contains the test set of the Phallusia mammillata dataset and used in the paper to train and validate the 3D Mask R-CNN. The test set is composed of the PM1 embryo images and ground truth. In details:

the 3D input image of Phallusia mammillata PM1 embryo (Inputs.zip)the ground truth ASTEC instance segmentation of these input images (ASTEC_Ground_truth.zip)the predictions infered by the trained 3D Mask R-CNN (3D_Mask_R_CNN.zip)the predictions infered by the state-of-the-art network for cell instance segmentation Cellpose with its pre-trained weights cyto3 (Cellpose_cyto3.zip)the predictions infered by Cellpose after being retrained over a sample of the Phallusia mammillata dataset (Cellpose_retrained.zip)the resulting weights of the 3D Mask R-CNN trained over the Phallusia mammillata dataset.These data can be used to reproduce the 3D Mask R-CNN inference over the Phallusia mammillata test set and to evaluate the predictions against the ASTEC ground truth.
The 3D Mask R-CNN code is available here: https://github.com/gdavid57/3d-mask-r-cnn
See the "morphogenesis" branch for the Phallusia mammillata dataset.