CASE STUDY

Bettering the eye care experience with machine learning

MACHINE LEARNINGIOS DEVELOPMENT

The Challenge

Pupillary distance (PD) and segment height (seg height) are two of the most important measurements required for buying prescription eyeglasses — and a common source of tension for optometrists and consumers alike. When someone receives a bad measurement, they get more than just a pair of glasses that doesn’t fit:

  • Wasted time + frustration
  • Wasted money from glasses they can’t use
  • Sore eyes and headaches
  • Long-term eye problems that can impact learning, overall vision development and eye health
Eyefit, a Des Moines optometrist with a goal of making the eyeglass fitting experience better, came to Shift with the question: How can we use technology to improve the measurement experience while ensuring “as good as the store” accuracy?

OUR
APPROACH

We started with a few weeks of in-depth discovery. In collaboration with Eyefit, Shift performed user interviews, technology research and competitive research. Knowing that other companies are using technology to measure PD and seg height, we dug into how we could make it better and more accurate. In our research process we found :

  • Online retailers are less focused on measurement technology; most don’t place significant emphasis on, or appear to have invested heavily into measurement tools
  • App-based companies appear to position themselves as disruptors in the process—going more all-in on customization at scale and/or how to shorten the gap from exam to purchase
  • Little mention of pain points in eye exam, measurement process, or health impact as it related to online glasses purchase showing a disconnect from customer needs

We also put together a quick prototype to ensure the technology side of things was even possible. We wanted to use Google’s Mediapipe to calculate pupillary distance and other measurements but were unsure of the accuracy. We developed a test utilizing Mediapipe and found it would work. After reporting these findings to Eyefit, we were on our way to development.

Eyefit approach 1
Eyefit approach 2
Eyefit approach 3
PreviousNext
Eyefit solution 1
Eyefit solution 2

The Solution

By utilizing machine learning with Mediapipe and Tensorflow, we created a white-label iOS mobile app designed for eye clinics and optometrists. A simplified app for in-home use calculates patient measurements quickly and easily. The application is built using Flask as a single endpoint that takes an input image and runs it through Mediapipe and Tensorflow, outputting a user’s pupillary distance and seg height.

After the initial build, we still wanted to make it more accurate. We utilized an industry solution, the credit card method, where a credit card is placed on your forehead so the application can more accurately pinpoint the parts of your face. By offering both options, the application is able to be even more accurate, gathering all of the data points an optometrist would need to accurately fit lenses.

TECHNOLOGIES

Mediapipe
Tensorflow
Machine Learning

SERVICES

Machine Learning
iOS Development

PROJECT
OUTCOME

Eyefit is still in the early stages of gathering funding but were excited to show stakeholders their idea was possible and get a test application out to begin using with prospective customers. Without the need to take in-office time and resources, Eyefit proves the possibility of expanding business opportunities for optometrists and eye clinics.

BIG IDEA?
WE CAN HELP
BRING
IT TO LIFE


LET'S CHAT

IMPROVING DEVELOPMENT READINESS THROUGH MVP VALIDATION


NEXT CASE STUDY

Have a question?

Give us a shout through our contact form below and we’ll follow up soon.

LET'S CHAT

Discover what's next.

Sign up to get a newsletter talking business and technology delivered fresh to your inbox.

Thrive on challenge?

We’re always looking to add talented individuals to the Shift team.

JOIN US