Published: Wed, January 17, 2018
Research | By Raquel Erickson

Cloud AutoML - Custom Machine Learning Models

Cloud AutoML - Custom Machine Learning Models

The key is that AutoML Vision will automatically train a machine-learning model for that data using well-known but complex techniques like learning to learn and transfer learning, said Jia Li, head of research and development for Cloud AI.

Cloud AutoML Vision will help users identify unlabeled images within a database of images when given a correctly labeled set of images to think over.

Cloud AutoML starts with image recognition, allowing customers to drag in images and subsequently task their operating systems to recognize the pics on Google's cloud.

"AI and machine learning is still a field with high barriers to entry and it requires expertise that (only) few companies can afford on their own", said Ms. Li, who is also the director of the Artificial Intelligence and Vision Labs at Stanford University.

This has previously seen the firm debut a mix of application programming interfaces (APIs), dubbed Vision, Speech, NLP, Translation and Dialogflow, that developers can combine with pre-trained machine learning models to advance the functionality of their business applications.

Google's first Cloud AutoML release is Cloud AutoML Vision, a service that makes it faster and easier to create custom ML models for image recognition. Since AutoML makes it possible to train a model in processing the specific type of data that a company wishes to analyze, the results can be more accurate than those produced by a general-purpose product such as the Vision API. Then voila - the AutoML system generates a trained model based on the user's own image data, complete with analysis and statistics showing the quality of the model.


Even though company says that AutoML is the only system of its kind on the market, the services like Clarif.ai and Microsoft's Cognitive Services can compete with Google. The technology was developed through collaboration with multiple other internal AI teams, she said.

Unless you've got a machine learning expert on the payroll, using those techniques to build the appropriate training model for your data set is extremely hard.

Those are some real-world implementations of Google's Cloud AutoML technology, which it launched globally on Wednesday (Jan 17). "The millions of images captured by these devices are then manually analysed and annotated with the relevant species, such as elephants, lions and giraffes, which is a labour-intensive and expensive process", she said.

The Zoological Society of London is using AutoML to provide automated detection of pictures of wild animals that are taken with a series of camera traps.

Google said the results from Cloud AutoML are more accurate than other "generic ML APIs", but it declined to get into details.

Like this: