Abstract: 

Azure Automated ML (AutoML) for Images allows to easily build and optimize computer vision models, without having to write any training code, while still maintaining complete control over model training, deployment and the end to end ML lifecycle of the model. It offers the following capabilities:

- Ability to optimize model performance by controlling model algorithms + hyperparameters
- Control over model training / deployment environment
- Ability to deploy the model to the cloud or download for local use
- Seamless integration with AzureML Data Labeling
- Operationalization at scale with Azure Machine Learning’s MLOps
- Support for Image classification, Object detection and Instance Segmentation

In this session, we will demonstrate how AutoML for Images can be used to create a computer vision model from your image data. You will also learn about the various advanced capabilities in AutoML like small object detection, incremental training, big data support using streaming and multi-gpu/multi-node training.

Bio: 

Radu is an engineering manager in Azure Machine Learning at Microsoft, where he works on AI infrastructure for Deep Learning. Most recently, he has been leading the team that develops Azure AutoML's computer vision capabilities - PyTorch deep learning models for image classification, object detection and segmentation. Prior to this, he designed and led the implementation of HyperDrive, Azure ML's distributed hyperparameter tuning system. In previous roles, Radu has worked on different projects, ranging from search engine infrastructure to information retrieval and data mining. He holds a PhD in compilers and programing language design from INRIA Nancy, France.