Once you've created a version of your dataset in Roboflow, you can export it directly to Microsoft Azure Custom Vision for training.
A dataset version in Roboflow:
A Custom Vision project in Microsoft Azure:
1. Select your project name in Microsoft Azure
2. Select API Key under "1" in order to take you to this page:
We'll copy the keys from this page to your Roboflow Workspace.
3. Navigate to your Roboflow Workspace's API keys in settings by clicking the dropdown menu n the Workspace.
You'll see "API Keys" as a sub menu:
4. Copy KEY 1 from your Microsoft Azure account to the "training-key" in Roboflow
5. Update the Endpoint field in Roboflow to match the Endpoint field from your Microsoft Custom Vision project.
From here, you're ready to start training with Roboflow and Microsoft Custom Azure together. Steps 1-5 only need to be completed once. To continue training with Azure, proceed.
6. Once you've generated a dataset version, export the project as "Azure Custom Vision."
The project will automatically upload to your Azure account, and you'll be able to proceed with training there.
(Note: ensure you're logged into your Microsoft Azure account in the same browser as your Roboflow account.)
Your exported dataset version will appear directly in your Azure account.
Once you've created a version of your dataset in Roboflow, you can export it directly to Google Cloud Platform Vision AutoML for training.
Dataset version in Roboflow:
Google Cloud Platform account
1. Export your dataset in the "Google Cloud AutoML" format
2. Copy the resulting link you're provided, and click "Continue to Cloud Vision." Ensure you're logged into Google Cloud Platform in the same browser as your Roboflow account.
3. Once in Google Cloud Platform, select "New Dataset," the type of dataset you exported from Roboflow (e.g. Object Detection) and proceed by clicking "Create Dataset"
4. Paste your link from Roboflow into the "Destination on Cloud Storage" bucket option.
From here, your dataset version is successfully imported to GCP Vision AutoML, and you can proceed with training in GCP.