Implementation:Huggingface Optimum ImageClassificationProcessing

Knowledge Sources	Huggingface_Optimum
Domains	Preprocessing, Vision
Last Updated	2026-02-15 00:00 GMT

Overview

Concrete tool for preprocessing image classification datasets with torchvision transforms and image processor normalization provided by the Huggingface Optimum library.

Description

ImageClassificationProcessing is a TaskProcessor subclass for image classification tasks. It accepts a BaseImageProcessor and builds a torchvision transform pipeline (Resize, CenterCrop, ToTensor, Normalize) using the image processor's size and normalization parameters. The default dataset is CIFAR-10.

Usage

Use this processor when benchmarking or evaluating image classification models. It handles image-to-tensor conversion with proper resizing and normalization matching the model's training configuration.

Code Reference

Source Location

• Repository: Huggingface_Optimum
• File: optimum/utils/preprocessing/image_classification.py
• Lines: 1-116

Signature

class ImageClassificationProcessing(TaskProcessor):
    ACCEPTED_PREPROCESSOR_CLASSES = (BaseImageProcessor,)
    DEFAULT_DATASET_ARGS = "uoft-cs/cifar10"
    DEFAULT_DATASET_DATA_KEYS = {"image": "img"}
    ALLOWED_DATA_KEY_NAMES = {"image"}
    DEFAULT_REF_KEYS = ["answers"]

    def __init__(
        self,
        config: "PretrainedConfig",
        preprocessor: "BaseImageProcessor",
        preprocessor_kwargs: Optional[Dict[str, Any]] = None,
    ): ...

    def dataset_processing_func(
        self, example: Dict[str, Any], data_keys: Dict[str, str], ref_keys: Optional[List[str]] = None
    ) -> Dict[str, Any]: ...

    def try_to_guess_data_keys(self, column_names: List[str]) -> Optional[Dict[str, str]]: ...
    def try_to_guess_ref_keys(self, column_names: List[str]) -> Optional[List[str]]: ...

Import

from optimum.utils.preprocessing.image_classification import ImageClassificationProcessing

I/O Contract

Inputs

Name	Type	Required	Description
config	PretrainedConfig	Yes	The model configuration
preprocessor	BaseImageProcessor	Yes	Image processor with size and normalization info
preprocessor_kwargs	Dict[str, Any]	No	Additional preprocessing keyword arguments

Outputs

Name	Type	Description
dataset_processing_func output	Dict	Adds "pixel_values" key as float32 numpy array

Usage Examples

from transformers import AutoConfig, AutoImageProcessor
from optimum.utils.preprocessing.image_classification import ImageClassificationProcessing

config = AutoConfig.from_pretrained("google/vit-base-patch16-224")
image_processor = AutoImageProcessor.from_pretrained("google/vit-base-patch16-224")

processor = ImageClassificationProcessing(config, image_processor)
dataset = processor.load_default_dataset(load_smallest_split=True, num_samples=50)

Related Pages

• Environment:Huggingface_Optimum_Python_Core_Dependencies