Implementation:Datajuicer Data juicer ImageDetectionYoloMapper

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

Overview

Concrete tool for performing object detection on images using YOLO provided by Data-Juicer.

Description

ImageDetectionYoloMapper is a mapper operator that uses a YOLO model to detect objects in images, returning bounding boxes and class labels. It loads a YOLO model (default: yolo11n.pt) via the ultralytics library and runs inference on each image with configurable image size, confidence threshold, and IoU threshold. Bounding boxes in xywh format and class labels are stored in the sample's metadata under bbox_tag and class_label_tag respectively. Requires CUDA acceleration.

Usage

Use when you need object detection annotations for images, providing spatial object information for downstream operators such as character detection and image analysis pipelines.

Code Reference

Source Location

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

Signature

@OPERATORS.register_module("image_detection_yolo_mapper")
class ImageDetectionYoloMapper(Mapper):
    def __init__(self,
                 imgsz=640,
                 conf=0.05,
                 iou=0.5,
                 model_path="yolo11n.pt",
                 *args, **kwargs):

Import

from data_juicer.ops.mapper.image_detection_yolo_mapper import ImageDetectionYoloMapper

I/O Contract

Inputs

Name	Type	Required	Description
imgsz	int	No	Resolution for image resizing, defaults to 640
conf	float	No	Confidence score threshold, defaults to 0.05
iou	float	No	IoU (Intersection over Union) score threshold, defaults to 0.5
model_path	str	No	Path to the YOLO model, defaults to "yolo11n.pt"

Outputs

Name	Type	Description
samples	Dict	Transformed samples with bbox_tag and class_label_tag in meta field

Usage Examples

process:
  - image_detection_yolo_mapper:
      imgsz: 640
      conf: 0.05
      iou: 0.5
      model_path: "yolo11n.pt"

Related Pages

• Environment:Datajuicer_Data_juicer_Python_Runtime_Environment