On January 30th Ken Taylor, our Senior Software Architect, gave a crash course into Machine Learning. All the information needed to participate were a functional knowledge of API (application program interface) and an attitude to learn and we’d like to share his key points of discussion.

How Can One Get Started with Machine Learning?
There are three ways to begin your machine learning journey; using a Cloud-based or Mobile API, using an existing model architecture and developing your own machine learning models for new problems. These listed methods are in order of complexity, so consider your existing knowledge base regarding machine learning and the effort you’re willing to put forth to try it on your own!

Cloud-based API’s

  • Cloud Vision can be used for faces, labels, OCR (optical character recognition), logos, safe search as well as landmarks and image properties.
  • Cloud Natural Language can be used for various language tools, such as syntax analysis, entity recognition and sentiment analysis.
  • Cloud Speech can be used for automatic speech recognition, global vocabulary, streaming recognition, inappropriate content filtering, real-time or buffered audio support, noisy audio handling and integrated API.
  • Cloud Translate can be used for visual text recognition for street signs and more; an example would be “translating” a stop sign in a foreign country to a language more familiar to you.

Common Mobile Vision API’s
Face API detects faces, facial landmarks, eyes open, and smiling. This can also be used in real-time to detect face bounding boxes and emotions.
Barcode API detects 1D and 2D barcodes, for example UPC and QR codes.
Text API detects Latin-based text and their structures. The text becomes segmented into blocks, lines and words.

What is TensorFlow?
TensorFlow is the most popular open source machine learning library and is especially useful for deep learning as well as research and production. Researchers use TensorFlow to devise new algorithms while developers use it to build new machine learning powered applications and services. Some exercises one can utilize from TensorFlow are MNIST for Beginners, MNIST for Experts, as well as other methods for each level of machine learners that can be found on their website.

Examples of Machine Learning
Deep Learning can be used to help understand languages, recognizing images with inception, play games such as chess and much more. In the arena of chess, the reigning champion Garry Kasparov was defeated by DeepBlue in 1997, the first computer to play at a championship level. Whether a novice or veteran in the arena of Machine Learning, we encourage you continue your venture into the study of these applications to help reinforce your wonder with it’s continuous innovation.

 

Google Developer Group

To find out more about Machine Learning in Southeast Virginia, check out the Google Developer Group of Hampton Roads.