Metadata-Version: 2.4
Name: msw
Version: 0.1.0
Summary: Multi-Scale Sliced Wasserstein distances and calibrated two-sample tests for stationary image distributions
Author-email: Youssef Haouchat <youssef.haouchat@epfl.ch>
License: MIT
Project-URL: Homepage, https://github.com/USERNAME/MSW
Keywords: sliced wasserstein,two-sample test,generative model evaluation,texture,stationary,image statistics
Classifier: Programming Language :: Python :: 3
Classifier: License :: OSI Approved :: MIT License
Classifier: Topic :: Scientific/Engineering :: Image Processing
Requires-Python: >=3.10
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: torch<2.7,>=2.0
Requires-Dist: numpy
Dynamic: license-file

# MSW

Multi-scale sliced Wasserstein two-sample tests for stationary image
distributions.

## Install

```bash
pip install git+https://github.com/HaouchatY/MSW.git
```

## Quickstart

```python
import torch, msw

A = torch.rand(64, 3, 32, 32)
B = torch.rand(64, 3, 32, 32)

r = msw.test(A, B)
print(r.T, r.p)                 # statistic, exact p-value
print(r.diagnose())

d, lo, hi = msw.distance(A, B)  # distance with 95% CI
```

More: `examples/quickstart.py`.

MIT license.
