Deep Learning for Food Recognition

Precision Agriculture
March 10, 2020
Self-Sovereign Identity
March 10, 2020

Deep Learning for Food Recognition

Artificial intelligence and deep learning now help food manufacturers and farmers determine nutritional deficiency and detect disease.

Key Insight

Deep learning is being used to identify food for a number of reasons: to help computers have more robust conversations with us about the content and origin of what we’re eating, to calculate the number of calories in a dish, and to spot spoiled or tainted food.

Why It Matters

Artificial intelligence and deep learning now help food manufacturers and farmers determine nutritional deficiency and detect disease. It helps consumers learn more about the provenance of our food.

Examples

Plantix, a cloud-based A.I. system, lets farmers identify pests and disease in their crops just by uploading photos of suspicious plants. The system will use image recognition to cross-reference with a database of various species, and within a couple minutes offer assessments of potential problems. Perhaps the plant is not getting enough water or needs a micronutrient.

California startup Abundant Robotics and Israeli-based FFRobotics are both developing automated picking systems that scan and “read” produce to determine when it’s ripe. SomaDetect lets dairy farms monitor milk quality using optical sensors and machine learning. Blue River Technology uses deep learning to automatically detect and spray weeds.

What’s Next

Deep learning will soon help determine exactly how much to feed livestock and will adjust quantities and mixtures of nutrients to optimize their health. Computer models will calculate the nutritional value of food before you’ve taken your first bite.

Researchers at the University of Massachusetts now use deep learning for computer-assisted dietary assessments, while scientists at Microsoft already incorporated their deep learning prototypes for recognizing popular Asian and Western foods into the Bing local search engine.

At the MIT Media Lab, students are at work on an organic barcode that’s invisible to us, but could be read by machines—it could be used to help consumers more easily trace produce as it moves around the world. Machine learning also lets chefs and at-home cooks determine which foods taste best together, select complementary ingredients, and offer food suggestions for various tastes.

The Impact

Deep learning can be used to find and sort problem products on food assembly lines, and it can help growers better identify crop disease. Deep learning for food recognition could soon present a number of opportunities for agricultural companies, farmers, food manufacturers, restaurants, chefs, and health-minded consumers.

Watchlist

Abundant Robotics, Alphabet, Apple, Blue River Technology, Carnegie Mellon, FFRobotics, IBM, John Deere Labs, Microsoft, MIT Media Lab, Penn State University, Plantix, PlantJammer, PlantVillage, Prospera, SomaDetect, University of Maryland, University of Massachusetts, University of Tokyo.