Implementation:Datajuicer Data juicer DetectMainCharacterMapper

Knowledge Sources	Datajuicer_Data_juicer
Domains	Data_Processing, Mapping
Last Updated	2026-02-14 16:00 GMT

Overview

Concrete tool for identifying and extracting main character names from images and their captions provided by Data-Juicer.

Description

DetectMainCharacterMapper is a mapper operator that uses a multimodal language model to identify main characters in an image based on both the image and its text description. It constructs a detailed prompt combining the image and caption, sends it to a multimodal LLM (default: LLaVA), and parses the JSON response to extract a list of main characters (people, animals, key objects) with their distinct characteristics. Samples with fewer characters than a configurable minimum threshold can be filtered out. Results are stored in the sample's metadata under main_character_list. Requires CUDA acceleration. It extends the Mapper base class.

Usage

Import when you need to identify main characters in images as the entry point for character analysis pipelines.

Code Reference

Source Location

• Repository: Datajuicer_Data_juicer
• File: data_juicer/ops/mapper/detect_main_character_mapper.py

Signature

@OPERATORS.register_module("detect_main_character_mapper")
class DetectMainCharacterMapper(Mapper):
    def __init__(self,
                 mllm_mapper_args: Optional[Dict] = {},
                 filter_min_character_num: int = 0,
                 *args, **kwargs):

Import

from data_juicer.ops.mapper.detect_main_character_mapper import DetectMainCharacterMapper

I/O Contract

Inputs

Name	Type	Required	Description
mllm_mapper_args	Optional[Dict]	No	Arguments for the multimodal language model mapper controlling caption generation. Default empty dict uses fixed values (max_new_tokens=256, temperature=0.2, hf_model=llava-hf/llava-v1.6-vicuna-7b-hf)
filter_min_character_num	int	No	Minimum number of main characters required; samples below this threshold are filtered out. Default: 0

Outputs

Name	Type	Description
samples	Dict	Transformed samples with main_character_list added to metadata containing character names and descriptions

Usage Examples

YAML Configuration

process:
  - detect_main_character_mapper:
      filter_min_character_num: 1

Related Pages

• Environment:Datajuicer_Data_juicer_Python_Runtime_Environment