FAQ

How much does it cost?

SnapCalorie is FREE for up to 3 AI logs per day!
There is also a premium subscription if you are really enjoying the experience and would like to take your health journey to the next level.

How accurate is SnapCalorie?

SnapCalorie is on average around 15% mean caloric error (that's +/- 150 calories on a 1000 calorie dish). By comparison nutrition labels are allowed to be 20% error, professional dietitians estimating from a photo are around 40% error, and average users of other nutrition tracking apps are 53% error.

What about oils?

The main place we recommend using the photo logging feature is when eating out. In these scenarios, perfectly measuring non-visible ingredients like oils, cooking fats, and sugar content would be impossible regardless of the approach or app used. Our app makes an educated guess based on average values for that type of dish and visual cues that it can see (oily vegetables or heavy salad dressing).

The main advantage of our approach is that it scans to volume of the food with your phone's depth sensor, something no other apps on the market do properly, and we give extremely accurate portion size. This is the main source of error for most people tracking, not oils and ingredients.

You can also add hints to the algorithm, such as the description from the menu, or anything you know about the ingredients, and our algorithm will automatically incorporate that into the prediction.

Why is SnapCalorie accurate?

Our team is comprised of several ex-Google AI researchers. In our prior research, we manually curated a dataset of 5,000 dishes where we weighed out every ingredient on the plate and then took hundreds of photos of these dishes from different angles and lighting conditions. These dishes included all of the hardest examples of food -- soups with ambiguous ingredients, sandwiches with heavy occlusions, dishes with oils and sugars, etc. We evaluated the accuracy of our algorithm in predicting the calorie and macronutrient contents of the dishes on this test set and measured how far off the predictions were on average.

This approach was peer reviewed by top researchers in computer vision and accepted to be published in the most prestigious academic conference in computer vision, CVPR. CVPR has an acceptance rate of around 25%, meaning around 75% of the papers submitted get rejected for not meeting the rigorous academic standards of this conference. This paper was co-authored by Perception Labs, Inc (the parent company of SnapCalorie) and Google Research. Read more here.

Can I use a wearable fitness tracker to deduct active calories?

You can integrate any fitness tracker with Apple Health and we will automatically pull active calories, workouts, and steps to offset your consumption calories with active calories burned throughout the day.

Support for Android integration currently in development.