Data Classes

Contents

Data Classes#

Anomalib’s dataclasses provide type-safe data containers with automatic validation. They support both PyTorch and NumPy backends for flexible data handling.

Generic Classes

Base dataclasses that define common data structures and validation logic:

  • Generic Item/Batch

  • Input/Output Fields

  • Validation Mixins

Generic Dataclasses
PyTorch Classes

PyTorch tensor-based implementations:

  • Image, Video, Depth Items

  • Batch Processing Support

  • Type-safe Validation

Torch Dataclasses
NumPy Classes

NumPy array-based implementations:

  • Efficient Data Processing

  • Array-based Containers

  • Conversion Utilities

Numpy Dataclasses

Documentation#

For detailed documentation and examples, see: