To create a labeling job, you must first prepare a dataset. Learn how to create datasets for data labeling and the required data formats.
Overview
Before labeling data in iTAG, create a dataset from the files to label. PAI supports two template types for labeling jobs. Data preparation and dataset creation vary by template:
-
Common template
Common templates support four data types: image, text, video, and audio. Steps and format requirements are covered in Create a text dataset and Create an image, video, or audio dataset.
-
Custom template
Custom templates let you label multiple data types, such as images and text, in a single job. Steps and format requirements are covered in Create a custom dataset.
Prerequisites
Object Storage Service (OSS) is activated. Get started with the OSS console.
Create a text dataset
|
Item |
Method 1: From cloud service |
Method 2: Local upload |
|
Procedure |
|
|
|
File name extension |
A .manifest or .txt file. |
A .csv or .xlsx file. |
|
File format |
The |
A column in the .csv or .xlsx file contains the text to be labeled. |
|
File example |
Create an image, video, or audio dataset
The following example uses images. The same procedure applies to video and audio files.
|
Item |
Method 1: Scan folder |
Method 2: Local upload |
|
Procedure |
|
|
|
File format |
The |
|
|
File example |
|
|
Create a custom dataset
|
Item |
From cloud service |
|
Procedure |
|
|
File name extension |
A .manifest or .txt file. |
|
File format |
Each line's For example, the following line defines a data item with an image stored at
|
|
File example |
Next steps
After creating a dataset, Create a labeling job.