update readme

t-vi · t-vi · commit 32ba4135c285 · 2021-03-07T21:38:09.000+01:00
diff --git a/README.md b/README.md
@@ -1,13 +1,57 @@
 # TorchDrift -- drift detection on PyTorch
 
+TorchDrift lets you monitor your PyTorch models to see if the data
+matches what it has been trained for. We focus on practical application
+and strive to seamlessly integrate with PyTorch.
 
+# Installation
+
+To install the latest release version use
+
+```
+pip install torchdrift
+```
+
+To get the latest and greatest install from git with
+```
+pip install git+https://github.com/torchdrift/torchdrift/
+```
+
+# Documentation
+
+Our documentation is at [TorchDrift.org](https://torchdrift.org/).
+
+# Examples
+
+Check out [our worked example](https://torchdrift.org/notebooks/drift_detection_on_images.html) with an ImageNet-type classifier.
+
+If you have a model (without head) as the feature extractor and a training dataloader you can fit the reference distribution as
+
+```python
+drift_detector = torchdrift.detectors.KernelMMDDriftDetector()
+torchdrift.utils.fit(train_dataloader, feature_extractor, drift_detector)
+```
+
+and then check drifts with
+
+```python
+features = feature_extractor(inputs)
+score = drift_detector(features)
+p_val = drift_detector.compute_p_value(features)
+
+if p_val < 0.01:
+    raise RuntimeError("Drifted Inputs")
+```
+
+Also check out our [deployment example](https://torchdrift.org/notebooks/deployment_monitoring_example.html) for integration of TorchDrift into
+inference with a model.
 
 # Authors
 
 TorchDrift is a joint project of Orobix Srl, Bergamo, Italy and
 MathInf GmbH, Garching b. M&uuml;nchen, Germany.
 
-Authors: ...
+The TorchDrift Team: Thomas Viehmann, Luca Antiga, Daniele Cortinovis, Lisa Lozza
 
 # Acknowledgements
 
