5. Export dataset

The full guide is part of Labelme Pro. See pricing.

Now you have done:

  1. Verification (Verify dataset visually, Verify dataset statistically)
  2. Additional data collection & annotation (Additional data collection)

so you're ready to move on to utilizing this dataset to train a model.

Convert dataset to a common format (masks)

To train a model (in this example dataset for object detection), you often want to use more common format of files (e.g., JPEG) other than the Labelme JSON format.


You can do that by using json-to-masks converting each annotation into a mask:

labelmetk json-to-masks amazon-picking-challenge-2016 -b

Browser view of json-to-masks output showing converted mask files for the dataset

Now you have done exporting the annotated JSON files into binary masks, which is more portable and compatible with other software.

If you have Labelme Pro, sign in to read the rest of the guide. Or see pricing.