AI meals scanner turns telephone images into dietary evaluation

March 18, 2025

The GIST Editors' notes

This text has been reviewed in accordance with Science X's editorial course of and insurance policies. Editors have highlighted the next attributes whereas making certain the content material's credibility:

fact-checked

trusted supply

proofread

AI meals scanner turns telephone images into dietary evaluation

Snap a photograph of your meal, and synthetic intelligence immediately tells you its calorie rely, fats content material, and dietary worth—no extra meals diaries or guesswork.

This futuristic situation is now a lot nearer to actuality, because of an AI system developed by NYU Tandon College of Engineering researchers that guarantees a brand new device for the tens of millions of people that need to handle their weight, diabetes and different diet-related well being situations.

The know-how, detailed in a paper offered on the sixth IEEE Worldwide Convention on Cellular Computing and Sustainable Informatics, makes use of superior deep-learning algorithms to acknowledge meals objects in photographs and calculate their dietary content material, together with energy, protein, carbohydrates and fats.

For over a decade, NYU's Fireplace Analysis Group, which incorporates the paper's lead writer Prabodh Panindre and co-author Sunil Kumar, has studied vital firefighter well being and operational challenges. A number of analysis research present that 73–88% of profession and 76–87% of volunteer firefighters are obese or overweight, dealing with elevated cardiovascular and different well being dangers that threaten operational readiness. These findings immediately motivated the event of their AI-powered food-tracking system.

"Conventional strategies of monitoring meals consumption rely closely on self-reporting, which is notoriously unreliable," stated Panindre, Affiliate Analysis Professor of NYU Tandon College of Engineering's Division of Mechanical Engineering. "Our system removes human error from the equation."

Regardless of the obvious simplicity of the idea, growing dependable meals recognition AI has stumped researchers for years. Earlier makes an attempt struggled with three elementary challenges that the NYU Tandon staff seems to have overcome.

"The sheer visible range of meals is staggering," stated Kumar, Professor of Mechanical Engineering at NYU Abu Dhabi and World Community Professor of Mechanical Engineering at NYU Tandon. "In contrast to manufactured objects with standardized appearances, the identical dish can look dramatically totally different based mostly on who ready it. A burger from one restaurant bears little resemblance to 1 from one other place, and selfmade variations add one other layer of complexity."

Earlier techniques additionally faltered when estimating portion sizes—an important consider dietary calculations. The NYU staff's advance is their volumetric computation operate, which makes use of superior picture processing to measure the precise space every meals occupies on a plate.

The system correlates the realm occupied by every meals merchandise with density and macronutrient information to transform 2D photographs into dietary assessments. This integration of volumetric computations with the AI mannequin allows exact evaluation with out handbook enter, fixing a longstanding problem in automated dietary monitoring.

The third main hurdle has been computational effectivity. Earlier fashions required an excessive amount of processing energy to be sensible for real-time use, typically necessitating cloud processing that launched delays and privateness issues.

The researchers used a strong image-recognition know-how referred to as YOLOv8 with ONNX Runtime (a device that helps AI packages run extra effectively) to construct a food-identification program that runs on a web site as a substitute of as a downloadable app, permitting folks to easily go to it utilizing their telephone's net browser to investigate meals and monitor their food plan.

When examined on a pizza slice, the system calculated 317 energy, 10 grams of protein, 40 grams of carbohydrates, and 13 grams of fats—dietary values that intently matched reference requirements. It carried out equally properly when analyzing extra complicated dishes comparable to idli sambhar, a South Indian specialty that includes steamed rice desserts with lentil stew, for which it calculated 221 energy, 7 grams of protein, 46 grams of carbohydrates and simply 1 gram of fats.

"Certainly one of our objectives was to make sure the system works throughout various cuisines and meals displays," stated Panindre. "We needed it to be as correct with a sizzling canine—280 energy in accordance with our system—as it’s with baklava, a Center Japanese pastry that our system identifies as having 310 energy and 18 grams of fats."

The researchers solved information challenges by combining related meals classes, eradicating meals sorts with too few examples, and giving further emphasis to sure meals throughout coaching. These methods helped refine their coaching dataset from numerous preliminary photographs to a extra balanced set of 95,000 cases throughout 214 meals classes.

The technical efficiency metrics are spectacular: the system achieved a imply Common Precision (mAP) rating of 0.7941 at an Intersection over Union (IoU) threshold of 0.5. For non-specialists, this implies the AI can precisely find and determine meals objects roughly 80% of the time, even once they overlap or are partially obscured.

The system has been deployed as an online software that works on cellular gadgets, making it doubtlessly accessible to anybody with a smartphone. The researchers describe their present system as a "proof-of-concept" that could possibly be refined and expanded for broader well being care purposes very quickly.

Along with Panindre and Kumar, the paper's authors are Praneeth Kumar Thummalapalli and Tanmay Mandal, each grasp's diploma college students in NYU Tandon's Division of Laptop Science and Engineering.

Extra info: Prabodh Panindre et al, Deep Studying Framework for Meals Merchandise Recognition and Diet Evaluation, 2025 sixth Worldwide Convention on Cellular Computing and Sustainable Informatics (ICMCSI) (2025). DOI: 10.1109/ICMCSI64620.2025.10883519

Offered by NYU Tandon College of Engineering Quotation: AI meals scanner turns telephone images into dietary evaluation (2025, March 18) retrieved 19 March 2025 from https://techxplore.com/information/2025-03-ai-food-scanner-photos-nutritional.html This doc is topic to copyright. Other than any honest dealing for the aim of personal examine or analysis, no half could also be reproduced with out the written permission. The content material is supplied for info functions solely.

Discover additional

What are macros? An train and diet scientist explains 9 shares