Presented By O'Reilly and Cloudera
Make Data Work
September 25–26, 2017: Training
September 26–28, 2017: Tutorials & Conference
New York, NY

Practical deep learning for understanding images

Leo Dirac (Amazon Web Services)
5:25pm6:05pm Wednesday, September 27, 2017
Artificial Intelligence, Machine Learning & Data Science
Location: 1A 12/14 Level: Intermediate
Secondary topics:  Cloud, Deep learning
Average rating: *****
(5.00, 6 ratings)

Who is this presentation for?

  • Software engineers, data scientists, and those aspiring to use machine learning

Prerequisite knowledge

  • A working knowledge of Python and MapReduce
  • Experience with Jupyter, PySpark, and basic image processing tools like PIL and ImageMagick (useful but not required)

What you'll learn

  • Learn how to apply the latest deep learning techniques (clustering, Apache MXNet, pretrained deep learning convolutional neural networks, and vector embeddings) to semantically understand images

Description

Deep learning techniques have revolutionized computer vision, making it possible to reliably understand what an image represents semantically, not just treat it as an array of pixels. Leo Dirac demonstrates how to apply the latest deep learning techniques to semantically understand images. You’ll learn what embeddings are, how to extract them from your images using deep convolutional neural networks (CNNs), and how they can be used to cluster and classify large datasets of images, as Leo walks you through a complete example, from preparing the images to visualizing the discovered clusters in a notebook. Along the way, you’ll see how deep learning networks produce vector embeddings as a side effect of their main classification task and how these can be used for related tasks such as measuring similarity.

Topics include:

  • Using Apache MXNet and Apache Spark to process images at scale
  • Using a pretrained convolutional neural network (CNN) to understand the content of an image
  • Using clustering to find groups of images with similar content
  • Using vector embeddings to summarize an image
Photo of Leo Dirac

Leo Dirac

Amazon Web Services

Leo Dirac is a principal engineer on the Amazon AI team at Amazon Web Services. Previously, he led the engineering team that launched the Amazon Machine Learning service. Leo has a background in physics. He started writing software professionally in the 1980s. In 2012, he became fascinated with deep learning and has been building systems with it ever since.