# Segmentation of nuclei using DSB-2018 top-1 neural network model

Based on [selimsef/dsb2018_topcoders](https://github.com/selimsef/dsb2018_topcoders/)

For comparison of Data Science Bowl 2018 best segmentation models see [Nucleus segmentation across imaging experiments: the 2018 Data Science Bowl, Juan C. Caicedo et al](https://www.nature.com/articles/s41592-019-0612-7).  

## Installation

1. Clone this repository

```

git clone https://github.com/yozhikoff/segmentation.git

```

2. Download [this](https://www.dropbox.com/s/qvtgbz0bnskn9wu/dsb2018_topcoders.zip?dl=0) and extract it to the

segmentation folder, replace all existing files using `Ay` keys when unzip asks about it. Note that you need to export to `/repo/segmentation/dsb2018_topcoders` withing the repo.

```

wget https://www.dropbox.com/s/qvtgbz0bnskn9wu/dsb2018_topcoders.zip?dl=1 dsb2018_topcoders.zip # note dl=1

unzip /path/to/zip/dsb2018_topcoders.zip -d /path/to/repo/segmentation/dsb2018_topcoders #type "Ay" when it asks about conflicts

```

3. Go to the segmentation folder and reset git files

```shell script

cd /path/to/repo/segmentation

git reset --hard

```

4. Create new conda env

``` 

conda create -n seg python=3.6.9 -y

conda activate seg

``` 

5) Install packages via conda and pip, simply (inside your conda env!)

```

sh ./install.sh

```

6) Test your installation using

```

python run_test.py

```

You can also try `example_notebook.ipynb` if you want to see usage details.