Iris tracking has been widely adopted in real-world applications such as computational photography and augmented reality. However, due to limited compute, changing light conditions and occlusions such as squinting or locks of hair, achieving accurate iris tracking on mobile devices is still a challenge.
A team of Google AI researchers has proposed a solution to this problem with MediaPipe Iris, a novel machine learning model designed to deliver accurate iris estimation without using depth sensors. Experiments show the approach can measure the distance from the camera lens to the user with a relative error rate comparable to methods that do use depth sensors.
Building on Google AI’s previous work with MediaPipe Face Mesh, the proposed model can track real-time landmarks involving the iris, pupil and the eye contours using a single RGB camera with no specialized hardware required. The approach is based on iris size, as human eyes’ horizontal iris diameter is mostly constant at 11.7±0.5 mm across a wide population.
MedaiPipe Iris was trained on cropped eye regions. The researchers manually annotated about 50k images that included various illumination conditions and head poses and ethnically diverse subjects. The resulting model can run on most modern mobile phones, desktops, laptops and even on the web.
To test model accuracy the researchers compared it to the depth sensor on an iPhone 11, collecting front-facing, synchronized video and depth images of over 200 participants. Using a laser ranging device, the team calculated the iPhone 11 depth sensor error rate as less than 2.0 percent for distances up to two meters. The proposed depth estimation approach had a respectable mean relative error of 4.3 percent and standard deviation of 2.4 percent.
The researchers stress that “any form of surveillance or identification is explicitly out of scope and not enabled by this technology” and that the model aligns with Google’s AI Principles.
They plan to extend the MediaPipe Iris model by improving stable tracking performance to achieve lower error rates, and hope the wider ML community can apply the its functionality in creative use cases, responsible new applications and new research avenues.
The MediaPipe Iris project page is on GitHub.
Analyst: Yuqing Li | Editor: Michael Sarazen; Yuan Yuan
Synced Report | A Survey of China’s Artificial Intelligence Solutions in Response to the COVID-19 Pandemic — 87 Case Studies from 700+ AI Vendors
This report offers a look at how China has leveraged artificial intelligence technologies in the battle against COVID-19. It is also available on Amazon Kindle. Along with this report, we also introduced a database covering additional 1428 artificial intelligence solutions from 12 pandemic scenarios.
Click here to find more reports from us.
We know you don’t want to miss any story. Subscribe to our popular Synced Global AI Weekly to get weekly AI updates.
Pingback: Google AI ‘MediaPipe Iris’ Improves Iris Tracking and Distance Estimation – Full-Stack Feed
Pingback: Google AI ‘MediaPipe Iris’ Improves Iris Tracking and Distance Estimation : artificial - Sixpost News
Pingback: Eyes on Me: Google AI ‘MediaPipe Iris’ Improves Iris Tracking and Distance Estimation – Tech Check News