Module preprocessing 

Shared preprocessing helpers for model adapters.

This module centralizes common preprocessing operations to reduce code duplication across model implementations. It provides:

• Configuration helpers for common model types
• Reusable preprocessing pipelines for common patterns
• Utility functions for image format conversions

Structs

DetectionResizer

Wrapper for DetResizeForTest to implement DetectionResizeOperation.

PreprocessPipelineBuilder

Builder for common preprocessing pipelines.

Traits

DetectionResizeOperation

Trait for detection resize operations that return scale information.

ResizeOperation

Trait for resize operations that return only resized images.

Functions

db_preprocess_for_text_type

Construct a DB preprocessing config based on text type.

db_preprocess_with_limit_side_len

Construct a DB preprocessing config that limits images by side length.

db_preprocess_with_resize_long

Construct a DB preprocessing config that resizes by long edge.

detection_resize_and_normalize

Applies a detection resizer (with scale info) and then normalizes the result.

pp_lcnet_preprocess

Construct a PP-LCNet preprocessing config with a custom input shape.

pp_lcnet_preprocess_with_norm

Construct a PP-LCNet preprocessing config with custom normalization stats.

resize_and_normalize

Applies a resizer and then normalizes the result to a tensor.

rgb_to_dynamic

Converts a batch of RGB images to dynamic images.

wrap_detection_resizer

Convenience function to wrap a DetResizeForTest for use in preprocessing pipelines.