BISSG

Code for paper IJCAI2022 "Biological Instance Segmentation with a Superpixel-Guided Graph"

Installaion

This code was implemented with Pytorch 1.0.1 (later versions may work), CUDA 9.0, Python 3.7.4 and Ubuntu 16.04.

If you have a Docker environment, we strongly recommend you to pull our image as follows:

docker pull registry.cn-hangzhou.aliyuncs.com/em_seg/v54_higra:9.1

Otherwise, please execute the following commands to ensure that the dependent software is installed

cd ./third_party/cython
python setup.py install
cd ../../
cd ./cython_function
python setup.py install
cd ../

Implementation

CUDA_VISIBLE_DEVICES=0  python -m torch.distributed.launch --nproc_per_node=1 train.py

Notice for CVPPP

This was officially confirmed by the authors of the dataset.

In the second half of 2021, the calculation of the SBD metric has been corrected for the bug that the SBD metric on the leaderboard site was higher than the `bestDice' metric. The previous SBD calculation on the website had a bug and it has been corrected recently https://codalab.lisn.upsaclay.fr/competitions/8970

The calculation method is corrected

from

SBD = np.amax([bestDice,bestDice0])

to

SBD = np.amin([bestDice,bestDice0])

Therefore, for a fair comparison, we use the previous result from the site https://competition.codalab.org as our results in Table 3.

Meanwhile, our best result has been updated on the leaderboard, as shown in the follows:

Contact

If you have any problem with the released code, please contact me by email ([email protected]).

Citation

@inproceedings{liu2022biological,
  title={Biological instance segmentation with a superpixel-guided graph},
  author={Liu, Xiaoyu and Huang, Wei and Zhang, Yueyi and Xiong, Zhiwei},
  year={2022},
  organization={IJCAI}
}