how to productionize custom vision model using Azure MLOPS?

Hi @Parthib Rathinam

Thanks for the question and using MS Q&A platform. 

In the model response, cannot see the bounding box co-ordinates details as json format, is there any API i can use to fetch these data in the response?

You can retrieve bounding box coordinates from your Azure Custom Vision model using the Prediction API. After sending an image for prediction, the API returns a response with the bounding box details. See: custom vision SDK. 

How to put this model into production and build end to end pipeline?

To productionize your Azure Custom Vision model, train and publish it, then integrate it using the Prediction API. Set up Azure Functions or App Service for handling requests, and use Azure Blob Storage for storing images. Automate workflows with Azure Logic Apps and implement CI/CD with Azure DevOps. Ensure scalability and secure access with auto-scaling and Azure Active Directory. This creates a robust and secure end-to-end pipeline for your model in production.

How to deploy this model on the edge device?

To deploy your Custom Vision model to an edge device, you can use Azure IoT Edge. Azure IoT Edge is a fully managed service that allows you to deploy and run cloud services, including machine learning models, on edge devices.

Please see: The edge with Custom Vision Service.

How to monitor the performance of deployed model , for example , show the dashboard which shows how many http requests came in, model accuracy, any data drift, server crash details etc.

To monitor the performance of an Azure Custom Vision deployed model, start by leveraging the Azure Portal to access performance metrics such as accuracy, precision, and recall through the "Performance" tab of your Custom Vision resource. Set up Azure Alerts to notify you of any anomalies or performance degradation. Regularly evaluate the model’s predictions with real or sample data and analyze usage patterns to understand how the model is performing under different conditions. By these approaches, you can effectively oversee and enhance the performance of your Custom Vision model.

See: Azure-monitor and azure-monitor_alerts.

I hope you understand. Thank you.