Tip
If you can't find answers to your questions in this FAQ, ask the Azure AI Vision API community on Stack Overflow or contact Help and Support on UserVoice.
Where is my key?
You can find your key in the Azure portal, under the key and endpoint tab.
Where is my prediction resource ID?
You can find the prediction resource ID on the resource's Properties tab in the Azure portal, listed as Resource ID.
Does my Custom Vision prediction endpoint accept base64-encoded images?
Yes. We recommend using open-source libraries that help you convert URL images to base64 encoded images and back. The OpenCV and Pillow libraries are popular with our customers for image conversion.
Is Custom Vision available in my region?
You can find the region availability for Custom Vision on the Region support page.
What is the difference between the free and standard pricing tiers?
See the Limits and quotas page for pricing tier information.
Do I have to relabel images if I change domains in a project?
You can switch domains in the project without relabeling, as long as you keep the same supervised learning domain (object detection or image classification).
Does Custom Vision support polygon bounding boxes in image tagging?
No, we only support rectangle bounding boxes.
I'm getting the message, `Can't train just yet. Reserved budget in hours is not enough to do advanced training`. What should I do?
If you get this message, you need to increase your training budget to train a higher volume of images.
My project training took many hours to complete. Why is the turnaround time so long?
High training traffic or high usage periods when many customers are training can affect training time. It leads to long queues and tight GPU resources. Sometimes, the job might stay in the queue and won't finish within a certain time frame, but it's guaranteed that our customers won't be charged for more than the set time budget.
Can I stop training once it has started?
There is no way to cancel a training once it has been started.
Once I train a Custom Vision model, can I manage/deploy the same model to different regions?
We don't have a use case for publishing a model to a different region, but we do offer the ability to export/import a project into different regions. See Copy and back up Custom Vision projects.
What is the difference between Custom Vision and AutoML?
Custom Vision is a tool for easily building Azure AI Vision models without needing to have any data science or ML knowledge. These models are pre-trained using datasets optimized for specific domains.
Sometimes, your scenario might require more control over model training, deployment, and the end-to-end ML lifecycle. AutoML is the ideal solution when you need this extra control, and it's addressed to customers with ML knowledge.
How can users be added to a multi-service account to collaborate on a Custom Vision project in the web portal?
You can use Azure role-based access control (RBAC) roles to give specific users access to collaborate on a custom vision portal project. See the Role-based access control docs.
Can training images be exported with the tags that were added in the Custom Vision portal?
Yes, you can use the GetImages API to export with tags.
How can I write logs on this service?
Use Diagnostic logging.