Skip to content

Latest commit

 

History

5 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 

Repository files navigation

2024 Tools for Image Annotation: task data preparation for Deep Learning training/finetuning

A quick lookup at 2024 image annotation tools: in 2024 one would expect to have a plethora of tools to label images for object detection/segmentation tasks, while this is somehow true, it is not immediately obvious which tool is better suited to which needs, hence this readme file.

Large Categories, pros and cons:

  • cloud based: + processing power, zero setup ; - data privacy, bandwidth
  • on premise: + data privacy, zero bandwidth; - requires setup
    • server side: + can access data on your remote workstation locally
    • client side: + can only access data on your local machine

Tested options

repo: complexity, cloud/premise, server/client, extra notes notes

  • HumanSignal/labelImg: intermediate, on premise, client side (executable)
  • ox-vgg/via: intermediate, on premise, client side, single html website which can be useful for local data
  • HumanSignal/label-studio: simple, on premise, server side, BUT sends lot of metrics and diagnostics and it's not currently possible to stop this behavoiur, requires account
  • l3p-cv/lost: complex, on premise, server side, (good for larger multi user companies)
  • UniversalDataTool/universal-data-tool: intermediate, can run as docker, premise, webapp, jupyter widget etc... but in practice it's not working any more
  • Hitachi-Automotive-And-Industry-Lab/semantic-segmentation-editor: bloated and overcomplicated, outdated
  • BMW-InnovationLab/BMW-Labeltool-Lite: intermediate, on premise, server side, the best tool to work with server-side data storage NOTE: ssh domain.com -L 8000:localhost:8000 -L frontendport:localhost:8081
  • RTIInternational/SMART: intermediate, dockerized, but not thought to manage data on the server side
  • IDEA-Research/deepdataspace: intermediate, client side, on premise (it's hard to find a tool thought to manage data on the server side!!!)
  • rediscovery-io/remo-python: looks amazing, but their website is down and both the pip package and the docker installation pull from 404 resources
  • AlvaroCavalcante/auto_annotate: ...
  • IndustryEssentials/label-free: mainly chinese, english full manual not existing, it has an option to load data from a server ip, but it's unclear
  • jaxony/turktool: simple, react app that can be used in many deployments fashion

About

A quick lookup at 2024 image annotation tools

Resources

Stars

0 stars

Watchers

1 watching

Forks

Releases

Packages

Contributors