datagym-ai/datagym-core

Open source annotation and labeling tool for image and video assets

DataGym.ai

DataGym.ai is a modern, web based workbench to label images and videos. It allows you to manage your projects and datasets, label data, control quality and build your own training data pipeline. With DataGym.ai´s API and Python SDK you can integrate it into your toolchain.

📒 Ressources

• Website: https://www.datagym.ai/
• Documentation: https://docs.datagym.ai/documentation/

🧩 Features

• Organize your data into different projects with tasks
  • Dashboard with useful statistics / overview
  • Tasks lifecycle with states (backlog, waiting, in progress, completed, skipped, reviewed)
  • Pagination, Filtering and Search
  • Integrated quality control / review process
• Organize your media within datasets
  • Different storage types (direct upload, public url´s, aws s3 cloud storage)
  • Supported mime types: jpeg, png, mp4
  • Support of large high resolution images
• Labeling features
  • Global classifications (image wide)
  • Image annotation
    • Variety of geometries: point, line, bounding box, polygons
    • Different classification types: text, checklists, option-box
    • Supports nested geometries (child-geometries)
  • Video annotation: Specialized editor for video labeling
    • Frame-by-frame navigation
    • Linear interpolation to track objects
    • Adjustable playback-speed
    • Analyze and extract video metadata (codec, framerate, duration, ...)
  • Image segmentation
    • Bitmap export
• Feature-rich Workspace
  • Temporary screen manipulations: contrast, brightness, saturation
  • Hide unused geometry-groups for more clarity
  • Shortcut support
  • Panning and zooming, multi-select, moving, duplication
  • Supports transformation of the same geometry type
  • Context menu for geometries
• Powerful REST API to build your own workflows
  • Python SDK Package
• Data exporting- and importing (json)
  • Export your labeled data as json (works for images and videos)
  • Import your labeled data to refine your ml model
  • Export-/import your label configuration and use it in multiple projects

🎯 Quickstart

Running with docker-compose

The simplest way to run DataGym.ai locally is by using docker-compose.

1. Download the docker-compose.yml from the projects root-directory

• https://raw.githubusercontent.com/datagym-ai/datagym-core/master/docker-compose.yml
• wget https://raw.githubusercontent.com/datagym-ai/datagym-core/master/docker-compose.yml

2. Launch container using docker-compose up -d
3. Wait until the initialization is done
4. Navigate to localhost:8080

Local development, build manually

Build the whole project:

mvn clean install 

🗳️ Build with

• Java / Spring Boot
• Angular

👐 Contributing

We would love to receive contributions - please review our Contributing Guide for all relevant details.

📜 License

This project is licensed under the MIT License - see the LICENSE file for details

datagym-core's Issues

Cant get project using the python api

Thanks for the amazing tool 💯

Issue Description

I encountered an error while running the import_coco_datagym.py script from the datagym-core package using Version: 0.7.2, Python 3.8.10. The error is related to JSON decoding and occurs during the execution of the script.

Steps to Reproduce

1. Run the import_coco_datagym.py script with the following command:

from datagym import Client

client = Client(api_key="d676b704-079f-4384-ae3f-XXXXX")

# projects = client.get_projects()
dummy_project = client.get_project_by_name(project_name="Dummy_Project")

Error Message

After running the localhost and obtaining the user token.

Traceback (most recent call last):
  File "datagym-core/import_coco_datagym.py", line 10, in <module>
    dummy_project = client.get_project_by_name(project_name="Dummy_Project")
  File "/home/pc-ws/.local/lib/python3.8/site-packages/datagym/client.py", line 192, in get_project_by_name
    projects = self.get_projects()
  File "/home/pc-ws/.local/lib/python3.8/site-packages/datagym/client.py", line 176, in get_projects
    projects = self._get_projects_without_images()
  File "/home/pc-ws/.local/lib/python3.8/site-packages/datagym/client.py", line 165, in _get_projects_without_images
    if self._response_valid(response):
  File "/home/pc-ws/.local/lib/python3.8/site-packages/datagym/client.py", line 129, in _response_valid
    elif json.loads(response.content)["key"] in WARNING_KEYS:
  File "/usr/lib/python3.8/json/__init__.py", line 357, in loads
    return _default_decoder.decode(s)
  File "/usr/lib/python3.8/json/decoder.py", line 340, in decode
    raise JSONDecodeError("Extra data", s, end)
json.decoder.JSONDecodeError: Extra data: line 1 column 5 (char 4)

The aws service is currently unavailable.

When I am trying to upload a .mp4 video manually through the interface, I receive a pop-up error message "The aws service is currently unavailable."

Invalid Docker-Compose.yml

[FR] Webhooks

Could the annotation server emit webhooks for certain events that external services can subscribe to? In particular I'd be interested in

• Annotation created
• Annotation updated
• Annotation deleted

need help

May I ask if there is a fee for me to use the collaboration feature? And why can't my account management page open? The following error is displayed: