Getting your data ready to train computer vision

Getting your data ready to train computer vision models using Azure Machine Learning

AI apps & agents a b Knowledge mining Machine learning

AI apps & agents a b Knowledge mining Machine learning

Azure Machine Learning Open & Interoperable Trusted Industry leading MLOps For all skill levels

Typical machine learning lifecycle Register Datasets Label Data (GA) (Preview) Train Models Deploy & Serve Monitor

Azure ML datasets

Managing data for ML is challenging Sources Environments Challenges Formats Insecure and fragile + + = Increasing storage costs Difficult to track & audit

Datasets : simple and secure data access

Dataset : access data through unified interface

Datasets : track data lineage in ML Training dataset ML experiment Test dataset Models

Azure ML datasets simplifies management Register Datasets Mount, Stream or download Load to common formats Simplifies access and management Secure & flexible + = + No more data copies Built in lineage tracking

Whacky data sets cheat sheet Tabular. Dataset File. Dataset

Demo

Azure ML data labeling

Azure ML data labeling Data Tooling Volume of Data Workflow Need to iterate Requires labeling tools Data movement Project Management Quality Control (human bias) Fundamentals Security Rate of labeling Distribution of labeling tasks Cost concerns Privacy

Demo flow Create dataset Create a labeling project Label data by self Add labelers Use labels to train a model Export labels as AML dataset Check quality of labels Labeling by other labelers

Demo

ML-assisted data labeling ml-assisted-labeling@microsoft. com

ML-assisted labeling dataflow (limited preview)

Clustered tasks

ML-assisted labeling dataflow (limited preview)

Pre-labeled tasks

ML-assisted labeling dataflow (limited preview)

Demo

Session Resources https: //ml. azure. com https: //aka. ms/ignite 2019 brk 3303 democode Datasets Creating data labeling project Labeling data ml-assisted-labeling@microsoft. com

Other session on Azure ML Day Time Code Title and speaker Monday, 11/04 4: 30 PM- 5: 15 PM BRK 2020 Creating a model factory with automated machine learning (Tzvi Keisar) Tuesday, 11/05 10: 30 AM-11: 15 AM BRK 3033 Top 10 things to know about Azure Machine Learning technical edition (Gregory Buehrer) 2: 15 PM- 3: 00 PM BRK 2019 Choosing the best language for your machine learning solution: From R to Python (Daniel Schneider) 3: 30 PM- 4: 15 PM BRK 2018 Responsible AI: Building trustworthy, secure and transparent machine learning (Sarah Bird) 9: 00 AM - 10: 15 AM BRK 1009 Machine learning simplified: From ideation to deployment with Azure Machine Learning (Erez Barak) 9: 15 AM - 10: 00 AM BRK 3035 Manage your end-to-end machine learning lifecycle with MLOps (Jordan Edwards) 1: 00 PM-1: 45 PM BRK 2017 Understanding enterprise readiness for machine learning solutions (Aashish Bhateja) 3: 40 PM - 4: 00 pm THR 3166 Data handling for the machine learning process using Azure Machine Learning datasets (May Hu, Dom Divakaruni) 1: 00 PM-1: 45 PM BRK 1007 Lessons learned from implementing real-world machine learning solutions: common patterns and best practices (Venky Veeraraghavan) 2: 15 PM- 3: 00 PM BRK 2184 Machine learning on Azure: Ask me anything (Venky Veeraraghavan) 11: 45 AM- 12: 30 PM BRK 2016 Azure Machine Learning and open source: Designed for each other (David Aronchick) Wednesday, 11/06 Thursday, 11/07 Friday, 11/08

Learn More c a b Start Free Documentation Give feedback Build, train, and deploy models with an Azure free account Dig into our technical documentation Tell us what you think, ask for a feature https: //azure. microsoft. com/free https: //aka. ms/Azure. MLDocs https: //aka. ms/Azure. ML_feedback

Invent with purpose.

Please evaluate this session Your feedback is important to us! https: //aka. ms/ignite. mobileapp https: //myignite. techcommunity. microsoft. com/evaluations

Find this session in Microsoft Tech Community