FAQs

A contributor to a project can see all the details, can annotate data items, can see the items done by others, download data but cannot add new data or add new contributors.

On the project details page, you can use the button "Add Contributor" to add a contributor by email Id.
They will receive a mail and when they log in, they will be able to see the project on their dashboard.

On the project details page, you can use the button "Add Contributor" to add an admin by email Id.
They will receive a mail and when they log in, they will be able to see the project on their dashboard.

For paid plans: You can delete your data anytime you wish.

On open data plans: We believe by using Dataturks services to build Open Datasets for your personal needs, you are contributing to the ML community at large by making the dataset open.

Also, for a fair application of the above belief, since these Datasets are open, once created you cannot delete the project or the data items in the project.

Yes, you can always revisit data items done and can make changes to them. You can filter data items by the annotator name or by the tagged labels.

We provide a summarized view of the items done by each annotator and the average time taken by them. Also, for each data item we provide the view on who annotated it and how long did they take to do the annotation.

API keys are allowed for paid plans and you can get them from your dashboard. API keys are per user and provide the same level of access as the user has on the UI.

Any tag you create on the data is counted as one label, example: one bounding box or one NER tag, so if in each image you draw 5 bounding boxes, then these are counted as 5 labels.

After your subscription expires, you can view and download data but can't upload more or edit annotations. So, read-only operations are allowed. You can always renew your subscription to get write access to it.

For paid plans (even during trials) all data is completely private. DataTurks claims no ownership of any of your data and never uses any part of your uploaded/tagged data for any internal use.

DataTurks Data security:

  • All your data is always private and never shared or exposed to any 3rd party.
  • At all points, your data belongs fully to you and Dataturks makes no claim of ownership whatsoever.
  • We never use the raw data or the tags you generate for any other internal or external use.
  • Internal access to the data is very restricted and only a few employees are allowed access to it.
  • You can delete your dataset whenever you wish, all deleted data is fully wiped out from our servers.
  • To provide fault tolerance, our systems are replicated on multiple cloud providers.
If you wish you can keep data on your internal servers and just upload URLs to them on Dataturks. You can also put IP restrictions etc such that these URLs are only accessible inside your office network.

A fully self-hosted data annotation solution. Takes 15 minutes to set up on servers, Works fully offline, no internet connection required, Docker-based installation supports all features as supported on the Dataturks cloud version.

Initially, only one account is added as the owner account for a subscription and all projects created from that account are private. You need to have your account also added as a co-owner to the subscription and then you too can create projects under the same subscription. Please mail from your registered email if you want us to add someone as a co-owner for the subscription.

You can upload raw images one-by-one or in one-go by uploading a zip file of all the images you want to upload.
Image URLs: You can also upload a text file where each line is a URL to the image.

All image formats which display on browsers like PNG, JPG, JPEG etc are supported, some special formats like TIF are only supported on Safari hence for these formats using Dataturks on Safari would work.

DataTurks download format is JSON, and the structure of the JSON varies slightly by task types. We provide handy scripts to convert this JSON to most popular formats like Pascal VOC, TensorFlow, MaskCNN.

DataTurks download format is JSON, we provide handy scripts to convert this JSON to Pascal VOC.

DataTurks download format is JSON, we provide handy scripts to convert this JSON to TFrecords.

DataTurks download format is JSON, we provide handy scripts to convert this JSON to masks.

You can upload raw files like text, PDF, Doc etc one-by-one or in one-go by uploading a zip file of all the files you want to upload.
File URLs: You can also upload a text file where each line is a URL to the text. Important, please read how to work with URLs to text files on Dataturks.

DataTurks download format is JSON, and the structure of the JSON varies slightly by task types. For NER, Document annotation and POS tagging we support downloading in Stanford NLP format.
We provide handy scripts to convert this JSON to popular formats like Spacy etc.

If you have a partially tagged data or you want to evaluate the output of your models, you can upload pre-tagged data for all of the tasks in DataTurks. The upload format of the pre-tagged data is almost always exactly the same as the download format from DataTurks. Please covert your pre-tagged data to DataTurks download format before uploading.

DataTurks download format is JSON, and the structure of the JSON varies slightly by task types. Please try downloading data from demo projects to check the download format.

Here is the example of the format for image tasks:

{"content": "http://image.com/img/a.png", "annotation": {"labels": [l1], "note": ""}, "extras": {"id": "39"}}

For each data item annotated, we return, the email of the annotator, the time taken to annotate, date-time of annotation, the current evaluation status of the item.

You can upload your data in the pre-tagged data format, with empty tags. Here you upload a JSON which supports adding extra metadata with each data item. Here is the example of the format for image tasks:

{"content": "http://myimage.com/img/abc.png", "extras": {"id": "39909", "hint": "shirt", "source": "you.com"}}