This topic describes how to create a labeling job.

Prerequisites

A dataset is registered. For more information about how to register a dataset, see Register a dataset.

Procedure

  1. Log on to the Machine Learning Platform for AI console.
  2. In the left-side navigation pane of the Machine Learning Platform for AI console, choose Data Preprocessing > Smart Labeling.
  3. On the Smart Labeling page, click Create Labeling Job.
  4. Enter the basic information, and click Next.
    Parameter Description
    Task Name The task name must be 1 to 24 characters in length, and can contain underscores (_) and hyphens (-). It must start with a letter, digit, or Chinese character.
    Description The description must be 0 to 64 characters in length, and can contain underscores (_) and hyphens (-). It must start with a letter, digit, or Chinese character.
    Input Dataset Select one or more datasets to create a labeling job. The datasets must correspond to the topic of the labeling job. If no dataset is available, click Register Dataset next to Input Dataset to register a dataset.
    Output Dataset Path Select an Object Storage Service (OSS) path to store the labeling results. When you handle a labeling job, every time you click Generate Result Dataset, a result dataset is generated in the specified OSS path. The dataset contains the results of all topics that you have completed.
  5. Click Next and set the parameters on the Template wizard page.
    Parameter Description
    Template The system provides the following templates: Target detection, Semantic segmentation, Image comprehensive labeling, OCR template, Single-label image classification, and Multi-label classification.
    Labels Labels are displayed in different colors. This parameter takes effect only when you choose Target detection, Semantic segmentation, or Image comprehensive labeling as the template.
    Label Image Orientation Enable or disable the feature of annotating image orientation. This parameter takes effect only when OCR template is selected as the template.
    Flip Text Enable or disable the feature of text flipping. If the text in an image is placed in the same direction as the image, you can turn off the switch. This parameter takes effect only when OCR template is selected as the template.
    Text Type Add labels that are used to classify the text. This parameter takes effect only when OCR template is selected as the template.
    Add Custom Label Add custom labels that are used to classify the text. This parameter takes effect only when OCR template is selected as the template.
    Categories Add labels that are used to classify images. This parameter takes effect only when Single-label image classification or Multi-label classification is selected as the template.
  6. Set the following parameters of the labeling policy and click Submit.
    Parameter Description
    Dispatch Policy The default dispatch policy is Number of topics collected by a worker each time and cannot be changed.
    Topics per Collection The number of topics collected by each worker each time.
    Note The specified number of Topics per Collection can be less than the total number of topics divided by the total number of workers. This allows workers with high efficiency to collect more topics and improves the overall efficiency of data labeling.
    Add Worker Add one or more workers. You can add both Alibaba Cloud accounts and Resource Access Management (RAM) users.