MASHINIST
Data Labeling Studio

Data labeling for training artificial intelligence models

Features

Labeling quality assessment
  • Quality control for each task and each user in real time
  • Automatic quality assessment
  • Timely identification of problems
Monitoring the performance of labelers
  • User’s rating = quantity + quality of tasks
  • Rating is reflected on a weekly basis
  • Rating may determine payment level and priority in the selection of tasks
Real time results analysis
  • Real-time task execution statistics.
  • Tracking the results of all labellers for each task
Task management
  • Curator can manage and view all his created tasks
  • The ability to create tasks, switch them on and off
  • View all tasks on one page
  • Biew task statistics and download input data and results
Results management

Exporting the labeling results in convenient machine-readable formats:

  • CSV
  • JSON
  • XLS
previous arrow
next arrow
Slider

Data Types

Task

Define the main sentiment of the text and choose the appropriate option

Sentiment analysis

Task

Select one from the proposed options and click confirm

Single choice

Task

Read the text. Click on one of the NER options, then select a portion of the
text

NER

Task

Select one of the two statements and press confirm

Pairwise comparison

Task

Read the text. Select one or several points from the offered options and
press confirm

Multiple choice

Task

Select one from the proposed options and click confirm

Validation

Task

Select one or more of the suggested options and click confirm

Multiple choice

Task

Select one from the proposed options and click confirm

Single choice

Task

Select a marker and mark key points on the image, define borders with
polygons

Polygons

Task

Select a marker and mark key points on the image

Landmarks

Task

Select a marker and mark the desired object on the image with a rectangle

Bounding boxes

Task

Listen to the audio, select one from the proposed options and press confirm

Single choice

Task

Click on the player button, listen to the audio and enter a description in
the ‘Description’ field

Free string

Task

Listen to the audio, mark the checkbox, and then mark an area on the track
field.

Segmentation

Markup on a turn-key

Random people working for pennies with low quality
scores

VS

Experienced professionals

+

Official work under contract

+

Compliance with privacy requirements

Do not entrust your data to random people

Labeling pipeline

Prepare your
data in JSON
Select a task type
Set a title and
description for the task
Download labeled
data in json /csv/xls
format
Activate task
Create guidelines
for labelers and set
up automated
quality control

Don’t trust your data to anyone?

Organize the labelling process with your team

Create a personal account for each curator and users
who report to them

Label data and enjoy automatic quality control

You can install a standalone studio module on the
company’s servers under a license agreement

Pricing

Purchase a license

  • On premise per server – 2 mln rubles
  • Cloud solution – 1 ruble per task

By installing Machinist studio on your server, you get the ability to label data without time
limit and the number of tasks

Turn-key labelling

  • Single choice – 7 rubles per task
  • Multiple choice – 12 rubles per task
Experts recommend labelling at least 1000 tasks per dataset, but no more than 50,000 tasks for one
type of object

Model Training Services

Dataset collection
Model training
Model Integration

Contact us