Implementation:Datajuicer Data juicer ImageShapeFilter

Knowledge Sources	Datajuicer_Data_juicer
Domains	Data_Quality, Filtering
Last Updated	2026-02-14 16:00 GMT

Overview

Concrete tool for filtering data samples based on image dimensions (width and height) provided by Data-Juicer.

Description

ImageShapeFilter is a filter operator that keeps samples with image shape (width, height) within specific ranges. It checks both the width and height of each image against configurable minimum and maximum thresholds. The image dimensions are stored under the image_width and image_height stats keys. The operator supports 'any' (keep if any image meets both width and height criteria) and 'all' (keep only if all images meet both criteria) strategies. It extends the Filter base class and implements the two-phase compute_stats/process pattern.

Usage

Import this operator when you need to filter dataset samples based on the pixel dimensions of images. Configure it in your Data-Juicer YAML config or instantiate directly.

Code Reference

Source Location

• Repository: Datajuicer_Data_juicer
• File: data_juicer/ops/filter/image_shape_filter.py
• Lines: 1-101

Signature

@OPERATORS.register_module("image_shape_filter")
@LOADED_IMAGES.register_module("image_shape_filter")
class ImageShapeFilter(Filter):
    def __init__(
        self,
        min_width: int = 1,
        max_width: int = sys.maxsize,
        min_height: int = 1,
        max_height: int = sys.maxsize,
        any_or_all: str = "any",
        *args,
        **kwargs,
    ):
        ...

Import

from data_juicer.ops.filter.image_shape_filter import ImageShapeFilter

I/O Contract

Inputs

Name	Type	Required	Description
min_width	int	No	The minimum width to keep samples. Default: 1
max_width	int	No	The maximum width to keep samples. Default: sys.maxsize
min_height	int	No	The minimum height to keep samples. Default: 1
max_height	int	No	The maximum height to keep samples. Default: sys.maxsize
any_or_all	str	No	Keep strategy: 'any' or 'all' across images. Default: "any"

Outputs

Name	Type	Description
samples	Dict	Filtered samples with stats field updated (image_width, image_height)

Usage Examples

YAML Configuration

process:
  - image_shape_filter:
      min_width: 100
      max_width: 4096
      min_height: 100
      max_height: 4096
      any_or_all: "any"

Python API

from data_juicer.ops.filter.image_shape_filter import ImageShapeFilter

op = ImageShapeFilter(min_width=100, max_width=4096, min_height=100, max_height=4096)
# Apply to dataset
result = dataset.process(op)

Related Pages

• Environment:Datajuicer_Data_juicer_Python_Runtime_Environment