Training data, algorithm tuning, and model testing for security

Why Figure Eight?

Facial recognition

There’s a very good reason that the most accurate facial recognition models succeed. That reason, simply put, is training data.

Think about on social media sites, where we tag our friends and loved ones with their real names. That’s all training data. It’s telling an algorithm who’s who and with enough of those decisions, the machine learning model can start to understand exactly that.

But not every company has access to billions of annotated photos. That’s where we come in. Our computer vision tool leverages human intelligence to annotate the facial features your algorithm needs to work and work well. That information allows you to build a corpus of known individuals that can be easily identified at any time.

Learn more

Computer vision for surveillance and monitoring

Facial recognition isn’t the only machine learning domain in security. Object recognition is also on the rise. Knowing that suspicious trucks appeared where you didn’t expect them or understanding movements that connote illicit behavior are things and AI can do, so long as it is shown enough examples.

We can do that. Check out our computer vision page to learn about our full capabilities.

Learn more

NDA Channels

We understand that security data often comes with privacy concerns. We offer customer-specific annotation schemas through special NDA channels that ensure your data is protected, secure, and accurately labeled at scale.

We’ll work with you, every step of the way, to make sure you get the data you need to train, test, and tune your machine learning security solution.

Learn more

Trusted by today’s leading brands

We needed a way to scalably measure search results quality. We needed to look at the impact of algorithms in search results relevancy and Figure Eight provides a way to do that. Our search results have gotten measurably better and we’ve been able to increase revenue, so it's been a real winner in terms of user experience and income.

James Rubenstein

Director of U.S. Search

Having tens of thousands of people at your disposal to read content and score it for sentiment and tone gives us the ability to inform our clients if there’s an issue. We can identify it quickly and put the campfire out before it becomes a forest fire.

Chris Lightner

Executive Vice President, Measurement and Insights

Machines don’t know music. Machines don’t feel the beat. Machines don’t feel happy. So we have to teach them. Teaching them involves two things: what do you put in, but also how do you grade your machines?

Henriette Cramer

Senior Research Lead



What We Learned Labeling 1 Million Images eBook

Read More

So you want to train a computer vision algorithm? Here are a few essential resources

Read More

Making a computer into a super-recognizer

Read More

See what Figure Eight can do for you