Hybrid Cloud

If you don't want Playment to access your data and only want to restrict access to your own annotators and team members, you can opt for the hybrid cloud approach.

In our usual approach, you host your raw data (images, point-clouds) in an s3 bucket and share it with Playment. Playment then copies the data over to its own bucket and serves it to annotators via signed URLs.

In the hybrid approach, you host your raw data in your AWS/GCP/Azure file storage or even your local file storage and don't share access with Playment. You then share just the URLs while creating jobs in Playment. You have to ensure that your annotators have access to objects in your file storage. This can be done by whitelisting the annotators' IP range if using AWS s3, GCS or Azure Blob storage or by adding your annotators in your local network.

You can also create a signed URL with an expiry date and send that signed URL as part of the job creation POST request.

To enable Hybrid cloud, you will have to make sure attachment validation is off in the workflow. You can go to the workflow settings and disable attachment validation.

Apart from raw data, there is another type of data, i.e. labels. These are stored either in Playment's database or Playment's s3 (eg. pixel-wise segmentation masks). In the hybrid approach, Playment will only store the labels, and not the raw data. Once a job is completed, You will be able to GET the labels via API and store it back in your file storage or database.

Last updated