New app SnapCalorie made by ex-Google engineer calculates calories in any meal by scanning a photo
>
Anyone who has dined out in recent years has felt the urge, and perhaps indulged, to snap a picture of a particularly well-plated or photogenic entrée for social media.
But now there may finally be a practical and legitimately constructive reason to shoot all those perfectly seared veggies, juicy burgers, or towering desserts with your phone.
Former engineers at Google and defense contractor Raytheon have created a cell phone app that can count the calories in any meal simply by snapping a photo.
Its makers say that SnapCalorie — which is already available for free download on Apple’s App Store and on Google Play, with a $29.00 per month premium option — is better at viewing a dish’s calorie content than “professional nutritionists’.
However, in a quick test at the office, DailyMail.com noted that SnapCalorie underestimated a side of broccoli on a lunch plate and included a photo of the vegetable, which was lightly buttered, as if only steamed.
“People are terrible at estimating the portion size of a plate of food visually,” said SnapCalorie’s co-founder, who hopes the app will find widespread acceptance among dieters who are too afraid to eat out in case they lose weight. exceed their calorie limit.
SnapCalorie’s founders say the app is better at looking at calorie content in a dish than “professional nutritionists.” The secret to their AI is a special dataset, Nutrition5k, which the company produced by collecting nutritional data, photos and videos from 5,000 real-world meals
A study compared the industry’s leading AI calorie-counting apps with dismal results, finding that SnapCalorie’s incumbent rival, Calorie Mama, was only right about 63 percent of the time
As of this month, Index Ventures, Y Combinator, and even CrossFit CEO Eric Roza himself have collectively invested a total of $2 million in the team’s new startup.
The funders have apparently seen something in the company’s co-founders, former Google AI engineer Wade Norris and ex-Raytheon engineer Scott Baron, as well as their calorie-counting AI, Roboflow.
But SnapCalorie’s real secret is its special dataset, Nutrition5k, which the company produced itself by taking nutritional data from 5,000 popular, real-world meals and pairing that data with thousands of photos and videos of those meals taken via a robotic rig.
Nutrition5k was used to train a computer vision algorithm to accurately guess the caloric content of everything from soups and burritos to trickier ingredients like oils and “mystery sauces.”
In an academic paper on training their algorithm with the Nutrition5k dataset, Wade and his team estimate that their AI’s average combined error for calories, food mass, and macronutrient mass is approximately 16.5 percent.
By comparison, their study found that the equivalent average error by a professional nutritionist was 41 percent and 53 percent for an ordinary person.
Prior to SnapCalorie, the company’s co-founder Wade Norris pioneered Google’s Google Lens project — another algorithm-driven computer vision tool, one that extracted public information about objects identified in the world.
With SnapCalorie, Norris said he was looking for a way to use that technology to more directly improve people’s lives.
Accuracy is very much the core goal of SnapCalorie, with its AI aided in portion size measurement not only by the modern phone camera’s unique depth sensors, but also by a team of expert human reviewers for “an extra layer of quality.”
“On average, the team is able to reduce the caloric error to less than 20%,” Norris said. TechCrunch this week.
In a quick test at the office, the DailyMail.com noted that SnapCalorie underestimated a side of broccoli on a healthy lunch plate, snapping a photo of the vegetable, which was lightly buttered because only steamed
“There are other apps that can use AI to track photo-based meals,” he noted, “but none of them help with portion size estimation — the most important part of reducing errors.”
SnapCalorie’s promises of unheard-of accuracy will have to be delivered if the makers are to make a name for themselves in the currently crowded marketplace: apps that use AI for calorie counting are already widely available to fitness-conscious consumers, including Lose It, Foodadviser, Bite.AI and Calorie Mama, though their results have been questioned by nutritionists.
A 2020 study in the Journal of Medical Internet Research’s Formative Research compared the industry’s leading AI calorie-counting apps with dismal results.
For example, Calorie Mama was only right about 63 percent of the time.
Currently, SnapCalorie results can be logged to the app’s native “food journal” or exported to popular fitness trackers like Apple Health.
At the current price point, users will likely want and expect that the app’s AI-generated calorie estimates will only get better and better with time.
Today, SnapCalorie’s premium model costs $29.00 per month, or nearly $200 cheaper per year at $149.00 per year.
The company’s recent $2 million cash infusion follows a previously raised $125,000 angel investment given by unidentified backers in what TechCrunch described as a pre-seed round.
But it remains an open question whether an automated photo capture and calorie counting app is more effective than the physical and emotional investment of a health journal.
A study published last year by Yale marketing professor Gal Zauberman and others found that people actually found the physical practice of manually tracking what they ate more effective and satisfying.
“This is a classic case of what behavioral researchers call ‘misprediction,'” Zauberman said of their study, published in the Journal of the Consumer Research Association.
‘People are drawn to the apparent ease of taking photos of their food’ he said. “But we notice that they follow less if they use the photo tool instead of the text tool.”