fast-stats is a fast and simple library for calculating basic statistics such as: precision, recall, and f1-score. The library also supports the calculation of confusion matrices. For examples, please look at the examples/ folder.
fast-stats is designed and tested against the scikit-learn library and provides formatted results to be a drop-in replacement in most cases. There are both generic functions like fast_stats.precision that provide significant speedup and specialized binary functions such as fast_stats.binary_precision that provide even faster optimal performance for binary data. fast-stats was specifically created to provide a limited alternative to scikit-learn.metrics for calculating statistics quickly on large multi-dimensional arrays or tensors such as those produced by computer vision models.
Benchmarks show approximately:
- 100x improvement in binary calculations
- 10x improvement in multiclass calculations
- 2x improvement in computing confusion matrices
- 15% speed-up over an equivalent
numpybinary calculation
The project was developed using the maturin framework.
See docs here: https://zachcoleman.github.io/fast-stats/
From PyPI:
pip install fast-statsBuild from source:
maturin build -r -i=path/to/python
pip install .../fast-stats/target/wheels/<whl file name>.whl
Tests are run with pytest.