Skip to content

CertifaiAI/classifai

Repository files navigation

GitHub GitHub release Sonar Cloud Quality Gate Documentation Discord

Data Annotation Platform for AI Training

Classifai is one of the most comprehensive open-source data annotation platform.
It supports the labelling of various data types with multi labelled outputs forms for AI model training.

Try it online:

Figure below show how Classifai fits in the machine learning workflow.
It enables the labelling of raw data imported from data source.
The labelled data can then channel into training environments for supervised / semi-supervised learning.

Data labelling tasks are challenging due to a few factors:

  • long hours of human workforce to label the data manually,
  • off-the-shelves toolkit which cannot fulfill the use cases needs.
  • frustration processes to convert data to supported format types for labelling work

We aim to solve these in Classifai by providing significant value to the data science workflow.

Features

  • Platform-agnostic
    • Classifai is build with Java backend.
      As Java is platform-neutral, it is as simple as download the classifai uber jar file and run it with Java runtime.
  • Asynchronous API - Fast and speedy response
  • In-memory Java-based database
  • Conversion of conventional data types to preferred formats (For example: .pdf format to .png format)
  • Multi data types supported
Images Documents Video Tabular Voice
jpg, JPEG, JPG pdf Coming soon... Coming soon... Coming soon...
png, PNG
tif, tiff
bmp
webp
  • [Stay tune with these features coming up below]
    • AI in the Loop - Deep Learning assistant for labelling task to reclaim valuable time for meaningful ML/DL tasks.
    • Support labelling of more data types in demand - video, tabular and voice data
    • Data management

Strengths of the tool

  • Scale data labelling operations to massive real world dataset
  • Cut costly data labelling services
  • Aesthetic and intuitive UI interface, to make the work fun to do!
  • Reclaim valuable time from inefficient data labelling, technical team can focus on more meaningful ML/DL tasks.

Quick Tour

Classifai support bounding box and segmentation annotation for now.
Click on the sub-category accordingly for the desired operation.

Bounding Box Annotation

Segmentation Annotation

Classifai is a web-based application which can opens in either Chrome, Firefox browser
or any Chromium-based browser such as Opera and Vivaldi.
(Note: Internet Explorer & Microsoft Edge is not supported)

There are two ways to open classifai in the browser

  1. Click on the first button of Welcome Launcher

  1. Alternatively, start classifai,
    then proceed to open a browser and put in url http://localhost:9999/

Conversion of files
We put into great thought into how data scientists build modelling with data.

When building use cases such as Optical Character Recognition (OCR) or medical related use cases,
often the raw data formats such as pdf/tif were not commonly used in the modelling.

Let alone data labellers were facing a hard time trying to convert these files into supported formats.

Classifai comes with a Conversion Launcher.
Currently supporting the conversion of format of pdf/tif to png/jpg.

Installation

Classifai supports the following Operating Systems.

Installation comes in distribution built with Java for each operating system.
The installation packages and formats are listed below.

Operating System Supported Version Installation package format
Windows 7, 8, 10 msi
Ubuntu 18 LTS, 20 LTS deb
Centos 7, 8 rpm
Mac 10 pkg

Alternatively, download the uber jar file and run with Java JDK/JRE 14.

java -jar classifai-uberjar-dev.jar --unlockdb --port=9999

Learn More

Section Description
Website Official Website
Documentation Full API documentation and tutorials
Blogs Technical Posts
Discord Community Support for Classifai Tool
AI from the Data Perspective Self-Paced Training to learn about Data Annotation and the use of Classifai

Contact us

For custom functionality development support, enterprise support and other related questions,
contact the team at [email protected]😃

BiWeekly Dev Meeting

Welcome to join! We have bi-weekly meeting to discuss about features, roadmaps and issues we are working on.