Skip to main content

Frequently Asked Questions

How Do I Upload DICOM data?

Please perform the following steps to register data that is stored on your S3 buckets with our platform:

  • Create a JSON file like this.
  • Follow the SDK tutorial to register all the files with our platform.

How Can I Set Up Upload Annotations?

You can find instructions for uploading the labels here in our documentation. We support JSON and COCO formats, but can help with writing a parser if you use a specific format.

How Can I Set Up Integrated Model Predictions?

We have multiple ways to integrate your model:

  1. Import your model prediction via the API
  2. You can upload the models and weights into our platform
  3. Custom integration - our platform can ping your models directly e.g. via webhooks, subject to Encord custom development effort.

Of these, 1 is the easiest way to start. You can upload the predictions of all of your models via our API.

Both 2 and 3 will require some engineering resources on your side.

How Do I Export Metadata Using Encord's SDK?

The review metadata will be exported in the JSON format so you will see the statuses e.g. what was reviewed and what wasn't. You find the annotation_task_status in the label_row that indicates the status - see statuses in our documentation here for more information.

How Does Team Management Work?

At an organizational level, you should have all the permissions that you need.

You are able to create new projects, ontologies (labelling structures), and upload datasets.

If you don’t have permissions on a specific project that someone else has created, you will need to be added to the project as an Admin (Project -> Settings -> Team) by the person who created the project.

If you want to have admin access to a specific ontology (labelling structure) or dataset that someone else created, you need to be added as an Admin as well.

The reason for having these granular access controls is to avoid accidental changes/deletions of data and labelling structures.

Download of labels via the SDK is slow. How do I speed it up?

If you try setting get_signed_url=False that should avoid the latency with signing the URL.

This can be done by using the following operation: project.get_label_row("<label_hash>", get_signed_url=False)

Here are the instructions for the get_label_row function.

How can I pre-select frames to label for a video via the SDK?

Add an extra feature in the ontology called, for example, Selected -> Yes (or whatever else you want to call it) and share the ontology tree with Encord Support.

This will allow us to match the import script with the feature uuid (featureNodeHash).

How do I avoid hitting the timeout when using the get_project() method on the API?

Increase the timeout of your Encord client, given the size of the project this is a long-running query.

Why can't I see all classifications/annotations in review the review workflow?

Check the sampling rate.

It's possible that you are seeing a different number of classifications/annotations than expected because you have the sampling rate set too high/low.

How can I get the label status via SDK?

To get the annotation_task_status you need to call the client.list_label_rows() function.

See our documentation for further details.

The possible response values will be QUEUED, ASSIGNED, IN_REVIEW, RETURNED, COMPLETED

Further information on annotation_task_status is available here.

How can I find ontology info in labels when exporting?

You can find ontology info in labels using encord_project.label_rows combined with encord_project.get_label_rows.

This yields a list of label rows that each has a list of labels.

Look for the object_answers and classification_answers entries, which can also be found in the label row.

If I have duplicate datasets, how can I find out which one is which?


  1. Go to datasets in GUI
  2. Re-name both datasets
  3. Open the cases in the editor and see which one is which (e.g. which one is annotated and which is not)
  4. Delete the one which is no longer wanted

How can I re-order an ontology?

This is possible using the SDK. The following example can be used to re-order the first two elements in an existing ontology:

  1. Authenticate the user client via the SDK using by defining the get_user_client function and invoking the EncordUserClient method if you haven't already:
def get_user_client() -> EncordUserClient:
ssh_key = os.environ.get("SSH_KEY")

return EncordUserClient.create_with_ssh_private_key(
  1. Pass the ONTOLOGY_RESOURCE_ID into the user_client's get_ontology method, perform the relevant changes, and call to commit the changes:
user_client = get_user_client()
ontology = user_client.get_ontology(ONTOLOGY_RESOURCE_ID)

first_object = ontology.structure.objects[0]
second_object = ontology.structure.objects[1]

ontology.structure.objects[0] = second_object
ontology.structure.objects[1] = first_object

N.B. If necessary, we can also do it manually for you in our database if you get stuck.

We will be looking to add a more user-friendly version of this functionality into our UI in the future.

How can I re-encode videos?

  • You can re-encode on our platform when uploading the data
  • For some encoding we need write permission on their bucket

Can I set up download via internet URLs?

We do not currently offer this because we prioritise the security of your data.

What are the currently-supported video formats?

We support a range of format and codecs:

  • .mp4 file with h.264 (not with h.265)
  • .webm withvp9

Others formats and codecs might be supported, but we cannot guarantee this.
We are willing to support your efforts with custom scripts where appropriate so please do reach out if you get stuck!

How do I set up the Azure integration?

To set up the Azure integration:

  1. Your User needs to be Admin of the Azure Tenant

  2. You need to add the application via the command line:

    New-AzADServicePrincipal -ApplicationId [APPLICATION_ID]

Please see our Azure documentation here for more information.

How can I import labels via CVAT importer?

Please see our CVAT documentation here.

What are Object Hashes in the SDK?

Each object hash is unique in the platform.

With DICOM images, I can show only one view. How do I correct this?

If this is happening, it is because some dimensions are different in the slices i.e. there is a corrupted slice.

All slices need to have the same dimensions in CT/MRI scans.

I am getting an error when I try to log in to the application. How can I fix this?

  1. Please ensure you're using the Chrome browser, and that it's up-to-date.
  2. Try clearing your browser cache & refreshing the page and trying again.
  3. If you're still stuck, please contact

I am getting an error when I try to display video/DICOM. How can I fix this?

This could be browser issue. To resolve, please:

  1. Use Chrome as a browser if you're not already
  2. Turn hardware acceleration off
  3. If you're still stuck, please contact

How can I change label status via the SDK?

You can change the labels status via the SDK or GUI using this guide.

This is a known issue with how links to the editor view on tasks currently work in our application.

We are looking to address this behaviour with an update soon.

At the moment, the URL leads to the editor view of the task, and if you haven't got edit permission on it, you will be redirected to the next available open task.

To solve this, just delete the last part of the URL /0/1 and reload the page. You will then be able to view the task in "view only" mode.

Are there any restrictions on the use of special characters in AWS S3 object names when importing as Encord private data?

Yes, in general it is advisable to avoid the use of special characters in object names.

See the AWS docs here for more information.

What does the "Version mismatch" pop-up in the label editor mean?

Don't worry, no data has been lost! This is a warning from our system that either two different people are editing a task at the same time, or that one individual has the task open in two different browser windows. The warning prevents annotators from overwriting others' or their own work.

In a scenario where someone else is editing the file, the warning pop-up will read as follows: "Version mismatch: Someone else is working on this. Are sure you would like to save your edit? This could overwrite changes that have been made."

In such cases, we recommend to click 'cancel' so as to prevent another annotator's work being overwritten. Then, ask your team who else might be editing the task at the same time.

In a scenario where the same user has the same task open in multiple browser tabs, the message will read as follows: "Version mismatch: The labels were updated outside of this session. Are sure you would like to save your edit? This could overwrite changes that have been made."

In such cases, we recommend to click 'cancel' to prevent any unexpected data overwrites and the annotator in question should perform an inventory on their browser tabs to discover where else the task is being edited.