PadimSharp brings the power of Padim (Patch Distribution Modeling) to the C# ecosystem.
Now you can train and detect anomalies entirely in C# – no Python needed!

🎯 Perfect for industrial inspection, quality control, and visual anomaly detection tasks.

• Pure C# implementation – Train and run inference seamlessly in .NET
• No external dependencies like Python or PyTorch
• Tested on MVTec – Proven performance on real-world anomaly detection benchmarks
• Built-in localization – See exactly where anomalies occur

// Train on normal images
Config config = new Config();
BaseModel model = new BaseModel(config);
model.Train();

// Detect anomalies on new images
(bool predictGood, torch.Tensor image) = model.Predict(predictImagePath);

• ✅ Train on custom datasets with normal samples only
• ✅ Fast inference suitable for real-time applications
• ✅ Pixel‑level anomaly localization
• ✅ Works with standard .NET image libraries (TorchSharp, etc.)

We've validated PadimSharp on the MVTec Anomaly Detection dataset – achieving reliable detection and localization across multiple object and texture categories.

Here's a real detection example from our MVTec tests:

Original Image	Predicted Anomaly Mask

The model successfully identifies the anomalous region with pixel-level precision.

We welcome contributions! Feel free to open issues or PRs to improve performance, documentation, or compatibility.

Say goodbye to Python‑based anomaly detection pipelines. Hello, PadimSharp. 🚀