This is an implementation of Mask RCNN for carrying out nuclei segmentation. The code has been referred from the Mask RCNN implementation by matterport.
The dataset for Kaggle Data Science Bowl 2018 has been used here. The dataset must be put into the folder samples/nucleus/data/.
Train a new model starting from ImageNet weights
cd samples/nucleus/
python3 nucleus.py train --dataset=data/ --subset=train --weights=imagenet
Train a new model starting from specific weights file
cd samples/nucleus/
python3 nucleus.py train --dataset=data/ --subset=train --weights=/path/to/weights.h5
Resume training a model that you had trained earlier
cd samples/nucleus/
python3 nucleus.py train --dataset=data/ --subset=train --weights=last
Generate submission file
cd samples
python3 nucleus_test.py
