Retirement and migration
Microsoft supports existing Azure Custom Vision customers until September 25, 2028. Its guidance identifies Azure Machine Learning AutoML and Foundry-based approaches as migration paths.
Microsoft retirement and migration guidance · checked October 8, 2026Options by requirement
| Option | Relevant capability | Team and implementation fit |
|---|---|---|
| StayCharted AI Model Trainer | Whole-image classification from labeled pictures. No object detection. Image classification → | Business teams wanting hosted training and file processing; developers can integrate API access on Essentials and up. |
| Azure Machine Learning AutoML | Custom image classification and object detection.Microsoft migration options · checked October 8, 2026 | Teams keeping workloads in Azure and managing ML deployment and integration. |
| Roboflow | Image classification and broader vision workflows.Roboflow vision tasks · checked October 8, 2026 | Computer-vision application teams. Compare workflows → |
| Amazon Rekognition Custom Labels | Custom image labels and object detection.AWS Custom Labels · checked October 8, 2026 | Teams building vision applications in AWS. |
| Google Cloud AutoML Image | Train image classifiers through the console or API; current Google documentation is under Gemini Enterprise Agent Platform.Google image classification · checked October 8, 2026 | Teams operating models within Google Cloud. |
| Azure Content Understanding (preview) | Managed custom classification of images and other unstructured content.Microsoft migration guidance · checked October 8, 2026 | Teams evaluating a managed Azure alternative; validate preview availability and requirements. |
Migration checklist
Microsoft’s Custom Vision migration guide
- Record the current categories, output format and application dependencies.
- Preserve the original labeled images and a separate test set.
- Train and evaluate the replacement, including rare categories and changed conditions.
- Update authentication, prediction calls, review handling and monitoring.
- Validate the complete workflow before retiring the old integration.
StayCharted plans and workflow
Free covers text and includes up to one AI Classifier Model. Essentials is $49/month and Business is $99/month per workspace. Pictures require Vision: +$49/month on Essentials or +$99/month on Business. Dedicated AI Models require Business. Prices are USD.
- Every training run reports accuracy on examples not used to train the model, overall and by category. Answers include confidence.
- Data-quality checks and personal-information masking before text training are available on every plan.
- On Essentials and Business, answers below your review cutoff go to a Review Queue. Grouping unlabeled examples into suggested categories is also included; pictures require Vision.
- Picture checks flag near-duplicates and potentially mislabeled examples. Location metadata is stripped from the copies used by the model; original uploads are retained unchanged until deletion.
StayCharted is a hosted service, with no on-premises application deployment or trained-model import. Standard workspace exports do not contain runnable models. Under Terms §6.3, an Owner can download a Dedicated model adapter after moving to a plan that no longer trains Dedicated models; running it elsewhere requires the exact base model and your own inference software.
Compare plans and allowances · Export conditions · Data handling
Common questions
When does Azure Custom Vision retire?
Microsoft says existing customers receive full support until September 25, 2028. Plan and validate a replacement before that support window ends.
Can StayCharted replace object detection?
No. StayCharted classifies whole images. If your application needs object coordinates or segmentation, evaluate a platform that supports those outputs.
Can I reuse my labeled pictures?
Prepare the original images and labels in the destination format, subject to your data rights and retention rules. Retrain and validate a new model; this page does not claim automatic transfer of Custom Vision model weights.
How should I compare migration options?
Keep a representative test set separate. Compare errors by category, rare-case recall, deployment requirements and total operating cost. Update the application integration and validate its outputs before switching.