Going Deep and Wide with A.I.

Our goal is to make our devices understand everyday objects fully — starting with sneakers!

SnapMe Sneakers

Learn more

SnapMe Sneakers A.I.

Neural networks and deep learning has made some incredible strides in many disciplines. We now have the possibility to go deep into specific domains and create systems that rival the best human experts. Our research team is applying deep learning to solve the various challenges product identification in images, videos and the real world.

It is the sheer scale of the world of products and the dynamics of ever changing product landscape that make it hard for even modern A.I. to keep track and stay up to date.

Adaptive systems

Plain similarity search will not solve this problem! The A.I. needs to learn real concepts how products are built, how they are configured and what they are for. Just as a human with all his experience and learned concepts can often discern products by just being shown one example each.

"The best deep neural network classifiers need weeks and large amounts of data to train — too slow and expensive for products and e-commerce!"

This adaptability is our goal. To be able to let our A.I. adapt to new product releases in realtime, learn their features and work efficiently with data it needs to learn the most abstract concepts just like humans do and gather as much general knowledge as possible. We are absolutely sure we will fry some hardware in the process!

"An A.I. that knows millions of products and can learn new products every hour. This is what we strive for."

Why Sneakers first?

We just love them! Besides, who doesn't know someone who asks "Where did you buy them?!" when you sport some new kicks?

The product categories is wide enough but also defined enough to be a great test bed for our technology and also bringing it to good use quickly. For shopping, for fun and also business analytics. — Nike? Adidas? Hear us? 😎

Similarity search? No!

We need Concepts, understanding and adaptability!

SnapMe Sneakers

Challenges!

Some of the challenges we are facing are so interesting that we'd like to share them with other fellow researchers and hackers.

Stay tuned for the upcoming challenges! You'll be able to win cash, sneakers or maybe a place in our team.

Challenge One
Using deep learning for quick finetuning

Sneakers Challenge
Using adversarial generative models to generate better product images

Want to learn more about what we do?

Sign up for the newsletter to get informed about our research progress and challenges.