Connections Introduction

📘

For Admins and Super Users

What are connections

Connections are a feature in Diffgram that helps you store credentials and meta data for other cloud/machine learning provider. Connection will help to seamlessly integrate across Diffgram's data pipeline on the stages of outputing results to external systems and receiving data from external providers and storage systems.

Why Use Connection in your Project?

  • Connect to cloud data, machine learning, and annotation providers with a 1 time setup.
  • Connect in standardized way. Easily manage a combine the data of your connections.
  • Manage multiple providers in a single place.

Connections will help in flowing data in and out of Diffgram with little techincal expertise. Just setups the credentials, and Diffgram will take care of the rest.

How Connections in Diffgram Work?

Diffgram makes sure to store your credentials in a secure way and makes sure to hook with the providers you give us in all the parts of the system that are necessary. From selecting files for data import to exporting directly to a cloud provider, Diffgram can automate the data extracion and preprocessing to have your files ready for labeling with little human interaction.

Connect to Cloud Data, Machine Learning, and Annotation Providers

Your data may reside in a cloud storage bucket. Perhaps you are training models with a cloud based engine or AutoML provider. Or an API backed external Annotation Provider. Connect these sources fast and securely with Sources.

Connect in Standardized Way

Sources are hand selected and implemented by the Diffgram team. This means you get an added degree of comfort when connecting. Standardized testing, analytics, and support. if your preferred Source is not included you can create a custom connection with the API or SDK.

Manage Multiple Providers

There are an increasing array of "best of breed" providers. You may have your primary storage on AWS, but want to train on Google's latest AI offering. Or want to supplement your in house team (ie outside of regular work hours) with annotations from ScaleAI for human in the loop.


What’s Next