Skip to main content

Export Annotations/Labels

How to download your annotations for use elsewhere

T
Written by Taylor Maggos
Updated over 8 months ago

So, you've finished annotating media in your Project and you want to take that data with you. Great! You can use these labels to visualize media in all sorts of other programs, such as geospatial data programs.

Annotations (also called data labels) can be exported directly from within the Project where they were created.

Steps to Export Labels

  1. First, go to the Annotate tab within your Project. Scroll down to the Annotation Tasks section and click Export Labels.

  2. Once you click Export Labels, you should see this message at the top of your screen appear in green --“Label export has been scheduled”.

  3. Now, you can switch back to the Projects Overview Tab, at the top of your page and monitor the Label Export box, midway through the page on the right-hand side.

  4. When your export is “Done”, click the download link to download to your computer.

Label Export Contents

Label exports will download a .zip archive containing the following files:

  1. A Readme text file which will give you information on what you just downloaded and how to view each file. **If this is your first time exporting labels from the platform, I suggest you read the Readme first!**

  2. Summary file (CSV) – a table of information corresponding to each task in the dataset including:

    1. Project ID (this will be the same for each task when you export labels from the same project)

    2. Dataset ID (this is helpful if you have multiple datasets attached to each project)

    3. Media ID (the unique ID we give to each piece of media you upload)

    4. Mask ID (the unique ID for each mask asset. Mask refers to the labels which have been annotate on each task, or outputted on each task by the model making predictions)

    5. Mask path (location of the mask within the platform)

    6. Title (original name of the image or video uploaded)

    7. Georeferenced (if the image or video is georeferenced or not)

    8. Label URL (link to the task in the annotation interface)

    9. Provider path (if the image or video has been uploaded from a specific provider, i.e. MAXAR, S3, GCP)

    10. Mask format (what format was the label annotated in, i.e. segmentation, detection)

    11. Created at (date and time of label creation)

    12. Category (how many instances of each category appear in that specific task)

  3. Masks Folder - a folder containing the exported labels (called "masks") in a JSON file format

Did this answer your question?