Image Recognition with AI “Fingerprints” • Sira Coba

(CEO & Founder, Shazura || CA USA & Spain)

“Since I was young, I knew I wanted to be a scientist. I was always imagining electrons flying around, waves in the air, visualizing the physics behind it. I’ve now been in science and artificial intelligence for almost 20 years, and am sure that will never change!

•

I’ve always been a very visual and imaginative person — that’s what laid the foundation for creating an image recognition algorithm and starting my company. After working for over a decade in image recognition for the military space, I had an idea while looking at a chair. I stood in a hotel staring at it and thought, ‘Why does my brain know that it is a chair when I haven’t seen this particular chair before?’ My brain doesn’t need to have seen it before to recognize it, nor train on hundreds of thousands of images of other chairs. It figures it out based on similar physical features to other related visuals. Moreover, it does it without the need of words or labels. The way the market currently does image recognition — training on huge datasets and labelling images — is actually exactly opposite to how humans do it. Typically, when we are born, it takes us two years to start talking, but we see from the first day. Imagine you’re a Martian and arrive here. You don’t know what’s a tree or a car, but you’ll still be able to differentiate between the two and recognize a subsequent tree without somebody telling you what it is, or needing to see thousands of trees before grasping what it is. So we created a unique mathematical model to represent any visuals, thinking of it as a ‘fingerprint.’ And that’s the story of how my company Shazura started.

•

I’m very happy with what we’ve achieved with the company, but it’s just the tip of the iceberg. My dream would be to help blind people see. So when I have time, I study neuroscience, so I can close the loop and take artificial intelligence to the human.”