Using facial images to determine BMI

Hi everyone,

I’m exploring how we can use facial images to determine (at least a correlation to) BMI and body fat percentage. Qualitatively I’ve been using my own facial image as a way of determining relative BMI / body fat percentage and as a self motivational tool. There has been some published research showing that this link is possible (face image to BMI) through machine vision and careful training of ML models.

I gave a presentation on the topic at the last New York Quantified Self show & tell, happy to share access to early prototypes and the deck if anyone is interested (video should be up in the next month or so). Starting to develop this into a more complete product, more information here -

How would you use a tool that derives BMI from facial images? (could be on yourself or others)

  • Shawn

Interesting idea! How do you deal with lens distortion and distortion due to different focal lengths?

In my case, an ML model might be able to better predict the BMI from the amount of facial hair :slight_smile:

Great question, I do agree that there will have to be some normalization work, I think the hardest part will be getting clean data. We initially want to look at Facebook pictures, but realize that many will have to be discarded due to quality. For now we’ll have to operate with the assumption that a lot of posted pictures are from a smartphone with a standard lens / distance from the camera. Would love to hear how you would tackle this issue, I see some interesting considerations in the article that you posted after reviewing briefly.

I don’t know if there is such a thing as a standard lens and distance, but perhaps (sun)glasses could help calibrate the measurements?

I am also interested in a Health Tracking tool for clients. We already have several important parameters, but this facial imaging tool sounds interesting not just for BMI tracking but overall eye clarity, facial color, clarity of skin, happiness, etc… I would love to learn more about it.

Hi there Uta, we are still in development, but would be great to chat about your interests, I’ll send you a DM to set something up. We’re looking at the power of self-visualization in general to drive behavior change and are looking across a number of parameters.

Also here is the video of my talk on some of our preliminary work, was just posted last week:

Hey Shawn - Do you have a link to any research? It’s an interesting concept and I’m intrigued to know how initial experiments have played out. I’ll admit, I’m a little bit skeptical (i.e people hold fat in different places) but always enjoy being proved wrong :slight_smile:

Nice talk BTW - love the timeline overlay against BMI. How selective were you for the photos used for analysis?

Would love to see the slides (on Slideshare) too.

Hey Dave, yes here is the link to the research - , and my slides -

I was not too selective in the photos of myself, just wanted to make sure I had a full on shot of my face available.

Sure I think there are instances of people holding fat in different places, but for the most part facial fat is a good correlate of overall fat %, of course one needs to account for race / gender / age / other compounding factors. Would welcome any research you have talking about the distribution of fat across people’s bodies, always looking to learn.