Google Calorie App: The Smart Way To Get In Shape

Google Calorie App: The Smart Way To Get In Shape

A Google calorie app, nicknamed Im2Calories, has recently been demonstrated at the Deep Learning Summit in Boston on the 26-27th May.

Primarily, the Google calorie app is a demonstration of technology, but nevertheless it gives great insight into what is just around the corner in the mobile world.


AI researcher, Kevin Murphy, works on experimental programs for Google and at the Boston summit presented his paper entitled ‘Semantic Image Segmentation Using Deep Learning & Graphical Models’.

Although the title gives little away to the layman, we are quite simply being introduced to a new revolution of smart artificial intelligence.

The software behind Im2Calories demonstrates its ability to recognise patterns in digital representations of sounds, images and data.

In this context, the Google calorie app looks at the pixels of a food image captured on a smartphone’s camera. It then uses this data to make a guess as to what the dish comprises of, how much of it there is and how many calories it contains.

Accuracy Rate

Google believes that if it can nail down the calories in your food snaps at a 30% clip then it will be onto a winner…

Im2Calories demonstrated its abilities by recognising a plate of two eggs, three strips of bacon and two pancakes. As these are not universally sized, the software powering the app judges the mass of each item in relation to the size of the plate.

The advantages of such an app are clear to see. It completely simplifies and automates the manual process of keeping a food diary by removing the guess work and manual input. Should the calorie counting app get confused as to what an item is, for example mistaking fried eggs with poached, then drop down menus can be used to correct the error.

It is believed that a 30% accuracy rate of the Google calorie app would be enough for consumers to get on board. The initial data collected by early adopters would be used to improve its accuracy over time and this leads us onto the papers main purpose.

Sophisticated deep learning algorithms

While this calorie counting app would be popular with consumers, it is the visual analysis process that opens the doors to a new breed of apps where AI is central to their design. By sensing the depth of the pixels in an image, married with pattern recognition and deep learning, we see the dawn of apps that will learn and improve themselves through their own use.

“The advantages of such an app are clear to see. It completely simplifies and automates the manual process of keeping a food diary.”

The whole point of this research is to develop a program where there is an absolute minimal amount of time spent by software engineers entering data. Achieving this will reduce both cost and time-to-market of new apps.

So, deep learning techniques aim to extract the meaning of imagery, text, audio and video to become somewhat self reliant. These techniques can then be applied to apps targeted at making any part of our daily life easier.

AI researcher, Kevin Murphy said “Suppose we did street scene analysis. We don’t want to just say there are cars in this intersection. That’s boring. We want to do things like localise cars, count the cars, get attributes of the cars, which way are they facing. Then we can do things like traffic scene analysis, predict where the most likely parking spot is. And since this is all learned from data, the technology is the same, you just change the data.”

Working in conjunction with the ever growing Internet of Things, the possibilities are endless. Data collected could be used for population level statistics for use in public health. Or, gardeners could take a snap of a pest devouring their plants and automatically be presented with a wiki on the pest, and how to eradicate it.

We have already seen deep learning algorithms in motion. The latest version of Google Now employs them to improve the accuracy of its own word recognition. In fact the deep learning update has already improved accuracy rates by 25%.

Deep learning is still relatively new and there is a lack of researchers working in this field. However, those prized individuals that do, are being snapped up by the likes of Google and Facebook. There is plenty of low hanging fruit to pick, so deep learning is about to launch the world into a new era of AI.

Don't forget to head over to for the most mouth-watering smartphone deals online.