What’s the blocker holding back computer vision AI (i.e. the technology that trains computers to interpret images and videos) from solving the hardest problems we face today?

According to Eric Landau, it’s the difficulty of creating usable, labeled, training datasets.

“That’s why we built Cord, a platform that can automate the data annotation process for computer vision use cases.”

We’ve briefly touched on the amount of data companies use to train AI models. Google is using all of your failed CAPTCHA attempts to train its AI and their latest Switch Transformer model (although an NLP) reportedly has been trained on 1.6 trillion parameters. The point: AI models need a lot of training data and makers can use help with annotation.

Cord automates the creation of training data through a “micro-model approach” so AI makers can build computer vision applications faster.

Landau explains how the micro-model tool works (summarized for readability):

1. Define the set of questions you are asking the data.
2. Do a handful of manual annotations to start teaching the system what you want to annotate.
3. Build a micro-model.
4. Run model, review the predictions, and retrain.

Sounds simple enough but this image helps for the non-technicals, too.

Landau says that Cord is already being used to help build datasets for computer vision applications that include the diagnosis of cancer, detection of people in danger in construction sites, and management of small business inventories.

The Product Hunt community is warmly welcoming Cord.

“Would've loved this in my previous gig where I was managing a team of computer visions experts who were always starving for properly annotated data sets.” - Kevin Simons

“Labelling training data is often the hardest part of ML projects. Glad to see someone tackling this part of the modelling pipeline!” - Elena Udodova

“Our ML team at Uber would have wished they something like this back in the days to label their data!” - Palle Broe

Take a look at the launch page to discover more potential verticals for Cord and how it differs from competitor CVAT in the comments.

• We’re giving away Product Hunt swag on Twitter. Shoot your shot.
• What’s your biggest learning experience from failure? Don’t keep it yourself.
• Google Sheets had its 15th birthday this week. These days you can make a website or CRM with it.
• The 2021 Apple Design Award Winners will be announced tomorrow. Congrats to Poolsuite, 1Password, Klima, and the rest of the finalists.