Skip to main content Module core Copy item path Source batch Batch processing utilities for the OCR pipeline. config Configuration management for the OCR pipeline. constants Constants used throughout the OCR pipeline. errors Error types for the OCR pipeline. image_reader Default implementation of the ImageReader trait used by inference builders. inference Structures and helpers for ONNX Runtime inference. macros Macros for the OCR pipeline. traits Trait definitions for the OCR pipeline. validation Input Validation Utilities AdapterInfo Information about a model adapter. AdapterTask A wrapper that implements Task for an adapter’s task type. BatchData Data structure for holding batched input data. BatchPerformanceMetrics Speed metrics for batch operations BatchSampler A sampler that creates batches of data with a specified batch size. CompatibleBatch A batch of compatible images that can be processed together CrossImageBatch Batch of cross-image items CrossImageItem Cross-image item for batching (e.g., text regions from multiple images) DefaultDynamicBatcher Default implementation of dynamic batcher DefaultImageReader Reads RGB images from disk with optional parallel batching. DynamicBatchConfig Configuration for dynamic batching DynamicBatchResult Result of dynamic batch processing ImageTaskInput Common input type for image-based tasks. ModelInferenceConfig Configuration for model inference and ONNX Runtime session setup. OpaqueError A simple, opaque error wrapper for cases where the original error type
is not Send or doesn’t implement std::error::Error. OrientationResult Represents an orientation angle with confidence information OrtInfer Core ONNX Runtime inference engine with support for pooling and configurable sessions. TaskRunner A task runner that executes tasks using a model adapter. TaskSchema Schema definition for task inputs and outputs. ToBatch A struct for converting image data into batched tensor format. TransformRegistry A registry for managing transform configurations. ConfigError Errors that can occur when validating configuration parameters. OCRError Enum representing various errors that can occur in the OCR pipeline. OwnedPredictionResult Enum representing owned prediction results. PaddingStrategy Strategy for padding images to uniform size PredictionResult Enum representing different types of prediction results. ProcessingStage Enum representing different stages of processing in the OCR pipeline. ShapeCompatibilityStrategy Strategy for determining shape compatibility TaskType Represents the type of OCR task being performed. TensorInput Represents a tensor input of various dimensions for generic inference. TensorOutput Generic tensor output supporting multiple data types. TransformConfig Configuration for different transform types. TransformType Enumeration of available transform types. DEFAULT_BATCH_SIZE The default number of items processed together in a batch. DEFAULT_CLASSIFICATION_INPUT_SHAPE The default shape (height, width) for classification images. DEFAULT_LIMIT_SIDE_LEN The default size to which image sides are limited during processing.
Matches the official PaddleOCR DBNet default of 960 (was previously 736). DEFAULT_MAX_IMG_WIDTH The default maximum width for images in the OCR pipeline. DEFAULT_MAX_SIDE_LIMIT The default maximum size for any side of an image. DEFAULT_PARALLEL_THRESHOLD The minimum number of items required before parallel processing is used. DEFAULT_REC_IMAGE_SHAPE The default shape (channels, height, width) for recognition images. DEFAULT_TOPK The default number of top results to select in classification tasks. MAX_TENSOR_SIZE The maximum number of elements allowed in a tensor, to prevent memory issues. AdapterBuilder Builder trait for creating model adapters. ConfigValidator A trait for validating configuration parameters. ConfigValidatorExt Extension trait for ConfigValidator that provides error wrapping utilities. DynamicBatcher Enhanced trait for dynamic batching functionality GranularImageReader Trait for image reading and I/O operations. ImageReader Trait for reading images. IntoOwnedPrediction Trait for converting a type into an owned prediction result. IntoPrediction Trait for converting a type into a prediction result. ModelAdapter Core trait for model adapters. Postprocessor Trait for postprocessing operations. Preprocessor Trait for image preprocessing operations. Sampler Trait for sampling data into batches. Task Core trait for OCR tasks. apply_document_orientation Rotates the image by the detected document orientation (0/90/180/270°);
unsupported angles return the image unchanged. apply_text_line_orientation Rotates the text-line image by the detected orientation (0/180°);
unsupported angles return the image unchanged. format_orientation_label Converts a numeric orientation label (e.g. “90”) to a degree string (e.g. “90°”). get_document_orientation_labels Gets the standard document orientation labels get_text_line_orientation_labels Gets the standard text line orientation labels load_session Loads a session with default logging configuration. parse_document_orientation Parses document orientation from classifier output parse_orientation_angle Parses an orientation angle from a label, accepting numeric (“0”, “90.0”)
and descriptive (“upright”, “inverted”, …) forms, validated against
valid_angles. Unknown or out-of-set labels yield an uncertain result. parse_text_line_orientation Parses text line orientation from classifier output validate_batch_size Validates that tensor batch size is positive. validate_division Validates division operands to prevent division by zero. validate_finite Validates that a float value is finite (not NaN or infinite). validate_image_dimensions Validates image dimensions. validate_index_bounds Validates that array index is within bounds. validate_non_empty Validates that a collection is not empty. validate_non_negative Validates that a value is non-negative (>= 0). validate_normalization_params Validates normalization parameters (mean and std). validate_positive Validates that a value is positive (> 0). validate_range Validates that a value is within a specified range (inclusive). validate_same_length Validates that two collections have the same length. validate_tensor_shape Validates tensor shape dimensions. OcrResult Convenient result alias for OCR operations.