Enabling Watson to see
Vision is one of the most powerful senses we have. As humans – most of us experience the world through our eyes. We are experts at instantly interpreting complex visual information into common language that we can share and understand. Vision is a powerful gift and a privilege, one that we often take for granted.
Today, it is my pleasure to introduce a major step forward– the ability for IBM Watson Visual Recognition beta service to learn from custom visual content. It’s now possible to train Visual Recognition to become an expert on custom trained natural scene images. Use Visual Recognition to identify your logo in social media, the difference between similar ships, or types of car models.
This expanded capability extends Visual Recognition to understand custom visual data, and can turn piles of images into organized information. The API recognizes what is in the images and returns text to describe what it sees.
To the average person this is a picture of a dog. If you asked 5 or 6 people who are dog-lovers they might be able to agree on a particular breed. However if you ask my sister – this is a picture of our lovable family mutt – Sarah. My sister is an expert at correctly identifying this dog specifically and distinguishing it from other dogs. Now, we can teach Watson to do the same. Visual Recognition can learn to identify and classify natural scene images so that the tag comes back as “Alyssa’s Dog ‘Sarah'”
Training Visual Recognition Learning API is easy – the service works best when you send a minimum of 50 positive and 50 negative image examples – the more the better. Check out the demo!
What will Watson see in your data?
[Note: AlchemyVision and Visual Recognition services will be merging capabilities in early 2016 – follow our blog to be notified for new offering announcements.]