This page contains the code of weakly supervised nuclei segmentation using points annotation proposed in [1].
Ubuntu 16.04
Pytorch 0.4.1
Gcc >= 4.9
GCC version >= 4.9 is required to build the CRF loss correctly.
cd ./crf_loss
python setup.py install- Put the original images in the folder ./data/MO/images and instance labels in the folder ./data/MO/labels_instance
- Specify the image names of train, val, test sets in the json file under ./data/MO
- Run the code:
python prepare_data.pyBefore training or testing the model, set the options and data transforms in options.py. Most options are set as default values,
and a part of them can be also parsed from the command line, for example:
python train.py --lr 0.0001 --epochs 60 --log-interval 30
python test.py --model-path ./experiments/MO/checkpoints/checkpoint_60.pth.tar
If you find this code helpful, please cite our work:
[1] Hui Qu, Pengxiang Wu, Qiaoying Huang, Jingru Yi, Gregory M. Riedlinger, Subhajyoti De and Dimitris N. Metaxas, "Weakly Supervised Deep Nuclei Segmentation using Points Annotation in Histopathology Images", In International Conference on Medical Imaging with Deep Learning (MIDL), 2019.