PyNeval is a Python package for evaluating the qualities of neuron reconstructions in the SWC format. It outputs quality scores of one or more test models by comparing them to a gold standard model. The scores also depend on which metric is specified because PyNeval supports several metric options.
Here is the simplest command line interface to run PyNeval:
pyneval --test <test_swc_path> --gold <gold_swc_path> --metric <metric>
test_swc_path is the file path to a test model, which is often produced by an automatic reconstruction method, and gold_swc_path is the file path to the gold standard model, which is typically created by manual editing. metric is the name of a quality metric, which can be
- length: Length metric for measuring the overlapping ratio between two models by matching line segments.
- ssd: SSD metric for measuring the overlapping ratio between two models by matching resampled nodes.
- diadem: DIADEM metric for measuring the amount of paths that contribute to topological similarity between two models.
- cn: Critical node metric for measuring the topological similarity between two models by matching topologically critical nodes.
pip install pyneval
conda install pyneval -c csdllab -c conda-forge
git clone https://github.com/bennieHan/pyNeval.git
cd PyNeval
python setup.py install
Once PyNeval is installed successfully, you can test it with the demo data included in the source code.
pyneval --gold data/demo/gold.swc --test data/demo/test.swc --metric ssd