Feature Engineering Methodes

Question

Hey There,

I am writing my Master Thesis at the moment. I am comparing AutoML products for image classification. There I compare the product Custom Vision with Google AutoML. However, I can't find the concrete methods of feature engineering and model selection from the documentation.

Thanks a lot!

Answer 1

@Arndt Bils Thanks for the question. Custom Vision is a tool for easily building computer vision models without needing to have any data science or ML knowledge. These models are pre-trained using datasets optimized for specific domains. Azure Custom Vision is an image recognition service that lets you build, deploy, and improve your own image identifier models. An image identifier applies labels (which represent classifications or objects) to images, according to their detected visual characteristics. Unlike the Computer Vision service, Custom Vision allows you to specify your own labels and train custom models to detect them.

AutoML is the ideal solution when you need this extra control, Here is the document to Set up AutoML to train computer vision models.

Azure ML customers can now easily build ML models trained on image data, without writing any training code. Customers can seamlessly integrate with Azure ML's Data Labeling capability to generate training data or bring their already labeled data. They can control which models to train by selecting from a variety of state-of-the-art algorithms and can optionally tune their hyperparameters to further optimize their performance. The resulting model can then be deployed as a web service in Azure ML or downloaded for local use and can be operationalized at scale by leveraging Azure ML’s MLOps capabilities.