This is a dataset markup & tagging tool
Now supports the following markup types:
- Text captcha, image classification markup
- Shapes on image (circle, rectangle, etc.)
- Points on image
Short summary:
Press Enter
to save text and go next
Use arrows (←
or →
) after typing Tab
or Esc
to switch between images
Shortcuts:
Alt+O
to open metadata file in default text editor
You need to select a folder with a dataset that contains numerically named directories with image.png
file inside (temporary format, I'll add support of different formats later)
After that program will count folders amount and show this. It will also create metadata.yml
file in the dataset directory (also there will be .metadata.yml
file when you start marking up your dataset - this is a "backup"). If the directory already contains a metadata file, DataTagger will read all necessary information. You can also see markup progress here. Finally, click "Go to tagging"
Markup interface is extremely simple. It shows you an image and you should enter text. You can click "Next" to save text and open the next picture. Also pressing Enter key does the same. If you need to go back or open another picture, press Esc
or Tab
, then you can use left and right arrows to open the previous or next image. There is also a shortcut Alt+O
to open the dataset metadata file in your default text editor for .yml
file
metadata.yml
has the following syntax:
!!me.veritaris.datatagger.Model.Metadata
datasetName: captchas
imagesAmount: 11259
lastTaggedImage: 1001
repeatedImages: [
]
taggedImages: {
}
Note: there are buttons "Finish tagging" and "Repetition", but they don't work for now. They will have functionality in the next release
Yes, I already know about ImageTagger, LabelImg, Yolo, etc. I just wanted to create a very simple, lightweight d x-platform app for datasets annotation for my needs. Why not if I can?