Imagine snapping a photo of your lunch and instantly knowing how many calories, carbs, and fat it contains—no need to guess or write anything down.
Thanks to a new AI system from NYU Tandon School of Engineering, this futuristic idea is now real and ready to use.
Researchers have developed a smart food scanner that uses artificial intelligence to recognize meals from photos and provide a full nutritional breakdown.
Just take a picture with your phone, and the system tells you the calorie count, protein, carbohydrates, and fat content in seconds.
This tool could help millions of people manage their weight, diabetes, or other diet-related health concerns.
It’s especially useful for those who find traditional food tracking methods—like keeping a food diary—too time-consuming or unreliable.
One of the key researchers, Dr. Prabodh Panindre, said, “Self-reporting what you eat is often inaccurate. Our AI system removes human error by doing all the work for you.”
The project was inspired by the team’s earlier research on firefighter health.
Many firefighters are overweight or obese, which increases the risk of heart problems. Helping them track their meals more easily could support better health and job performance.
So how does the system work?
First, it uses a powerful image recognition tool called YOLOv8, which spots different types of food in a picture—even if the food is overlapping or partly hidden.
The AI was trained with a carefully chosen set of 95,000 images, covering 214 types of food from around the world. The team made sure the system could recognize both simple and complex dishes—from pizza and hot dogs to idli sambhar and baklava.
Next, the system estimates portion sizes using a smart image processing technique that calculates how much space each food takes up on the plate. It then uses that information, along with known food density and nutrient data, to give accurate results.
Unlike older systems that were slow and required cloud computing, this one runs directly on a mobile website. You don’t need to download an app—just open the website on your phone and start snapping photos of your meals.
Tests show that the AI is around 80% accurate at identifying foods and estimating their nutrition. For example, it calculated a slice of pizza as having 317 calories and a hot dog as 280—very close to official values.
While it’s still a prototype, the researchers believe this tool could soon be expanded for everyday health tracking and even used by healthcare providers.
“This is just the beginning,” said co-researcher Dr. Sunil Kumar. “Our goal is to make healthy eating easier and smarter for everyone.”
