How Deep Learning Will Stir More Joy into Your Cooking

We all need all the help we can get in the kitchen. Whether that involves meal planning or roasting the perfect chicken, deep learning promises to help bring more to the joy of cooking.

Thanks to deep learning, you’ll soon get the kind of help no cookbook can provide. Innit, a Bay Area startup, is tapping the power of GPU-powered deep learning to build what can best described as a kitchen operating system.

“It’s all about you and your food and the appliances you use every day, and the people you share your food with,” Hristo Bojinov, Innit’s CTO, said during a presentation Thursday at NVIDIA’s GPU Technology Conference. “We like to say that we’re giving food a voice.”

Innit’s vision combines connectivity and computer vision to make the most of food in our refrigerators and pantries. It’s building its platform with the intent of tapping multiple data sources—shopping records, a recipe database, market analytics, and user behavior and preferences—to streamline meal preparation.

Bojinov said the company wants consumers to be able to spend more time with friends and family, and less time planning meals, shopping, and monitoring the cooking process.

It starts with taking advantage of the concept of the connected kitchen. Bojinov said today’s “smart kitchens” aren’t so smart. But “they’re getting increasingly connected, and that’s a trend we leverage in our work.”

Innit aims to provide capability to monitor a kitchen’s inventory, use shopping records to know when items will spoil, match supplies with recipes to identify best options, and coordinate the operation of appliances to simplify preparing those recipes.

At the core of this is computer vision powered by deep learning.

“We view computer vision as something that not only acts on the products you have, but also on cooking,” said Bojinov. “This is kind of a dynamic capability.”

Innit’s core “food recognition service” acts as an orchestration engine, and it uses a backend instance of Amazon Web Services’ G2, running a blend of GPUs and CPUs, to power its processing. The company’s computer vision pipeline relies on a deep neural network for object detection and classification, and GPUs handle feature extraction during the product matching process.

“The GPU has been a big enabler both in terms of being able to serve results quicker for our users, and also for price point,” said Bojinov.

For more, tune in to our interview with Bojinov, live from the floor of this week’s GPU Technology Conference.

 

AI Podcast: Food Delivery Bot Hits the Streets of San Francisco

And if you missed our podcast last week, it’s worth a listen: We spoke with the team at Marble, which has turned AI loose on the streets of San Francisco to deliver food in San Francisco’s Mission District.

Featured Image Credit: Jameel Winter, via Flickr

How to Tune in to the AI Podcast

Our AI Podcast is available through iTunes, Google Play Music, DoggCatcher, Overcast, Podbay, Pocket Casts, PodCruncher, PodKicker, Stitcher and Soundcloud. If your favorite isn’t listed here, leave a comment below.

Listen on Google Play Music itunes

Similar Stories