Researchers Compile 20 Million Images of Eyes in a Massive Open-Source Dataset

Countless modern applications use eye-tracking software or rely on the user’s eyes. Whether it’s for AR/VR gaming or for helping people with disabilities, there’s much to be done with eye data. Enter TEyeD, a public data set with over 20 million images of eyes collected from a team of researchers at University Tübingen, Germany, who hope to see some good come from possibilities of eye-tracking devices in their paper, “TEyeD: Over 20 million real-world eye images with Pupil, Eyelid, and Iris 2D and 3D Segmentations, 2D and 3D Landmarks, 3D Eyeball, Gaze Vector, and Eye Movement Types.”

The Method:

The TEyeD data set was created using seven different head-mounted eye trackers across a variety of situations. Two of them were placed into virtual reality and augmented reality devices.

The 20 million images weren’t just taken in a lab at various angles, rather, they were obtained through a variety of tasks, such as in car rides, simulator rides, outdoor sports activities, and daily indoor activities. The data set includes 2D\&3D landmarks, semantic segmentation, 3D eyeball annotation, and the gaze vector and eye movement types for all images. For the videos, they range from mere seconds to multiple hours.

The Context:

Many applications and devices use eye-tracking software, so this open-source dataset, TEyeD, will be a welcome (and massive) boon to other researchers. Human eye movements affect so much – more than just what we’re looking at right now. Some devices can be used for fun, such as with AR and VR games, while other applications can be life-saving, such as with tools to help those with disabilities.

The dataset is available here.

More on eye tracking, healthcare, and unique use cases beyond TEyeD:

This research is certainly insightful, but it doesn’t stand on its own. To make use of it, you need to know more about use cases, human-computer interaction, and the various disciplines where this data might be useful. Here are some Ai+ Training sessions that may prove valuable:

Playlist: Ai and Healthcare, Life Science, Pharma, and Biotech

Introduction To Face Processing With Computer Vision

Deep Transfer Learning for Computer Vision: Real-World Applications at Nanoscale

There are also a few relevant talks coming to ODSC East 2021 this March 30th-April 1st that may prove useful as well:

AI-Powered Best Healthcare for Everyone

The Clinician’s AI Partner: Augmenting Clinician Capabilities Across the Spectrum of Healthcare

Computer Vision for Healthcare and Medicine