Batches (beta)

📘

Beta

Batch queue is still in beta

The Batch Queue feature allows you to select individual Data Rows from any Dataset on the platform and submit them to a Project for labeling.

You can switch between the new Batch Queue and standard Attached datasets at any time. When you activate Batch labeling, you can think of the Data Rows you send to a Project as a separate queue. When your project is in Batch selection mode, Labelbox dedupes by Data Row to prevent duplicate labeling from the Dataset Queue.

Change queue mode

Projects by default are in Dataset mode. You can change the queue mode using the SDK or the App.

App

To activate Batch selection for your project, select an existing Project, go to the Labels tab, click on the Queue subtab, and select Use batch selection.

SDK

from labelbox import Client
from labelbox.schema.project import QueueMode

client = Client()

project = client.get_project("<project-id>")

project.queue_mode # Dataset by default 
project.update(queue_mode=QueueMode.Batch)

Select Data Rows

App

After you activate Batch selection for your Project, follow these steps to select your Data Rows to label:

  1. Navigate to the Catalog tab to select one or more Data Rows to label.

  2. After you select your Data Rows, you will see a dropdown menu with an option to Add batch to project.

  1. Review the selected Data Rows and select a Project to send them for labeling.

  2. Click Add to project.

  1. Finally, navigate to the project to begin labeling. Reminder, make sure the project is Batch mode to access these new datarows.

📘

Note

Newly queued Data Rows will be distributed after any Data Rows in the label queue that have already been reserved.

SDK

You can queue and dequeue data rows from a project using the SDK. Queue and dequeue accept data row ids.

import numpy as np

sample_size = 100
dataset = client.get_dataset("<dataset-id>")
datarows = list(dataset.export_data_rows())

sample = np.random.choice(datarows, sample_size, replace=False)
sample = [dr.uid for dr in sample]  # upload by data row id

# queue sample data rows
project.queue(sample)

# remove some data rows
project.dequeue(sample[:20])

What’s Next

Select data with Catalog

Did this page help you?