Thursday 20 March 2025
The quest for high-frequency eye tracking has been an ongoing challenge in the field of computer vision and machine learning. For years, researchers have been working towards developing systems that can accurately track eye movements at speeds of over 10,000 Hz. This may seem like a niche problem, but it’s crucial for applications such as augmented reality, virtual reality, and even medical research.
Recently, a team of researchers from the EssilorLuxottica Smart Eyewear Lab has made significant progress in this area by developing an improved version of an existing dataset for event-based eye tracking. The dataset, originally proposed by Angelopoulos et al., is one of the few available datasets that uses event cameras to capture eye movements.
Event cameras are a type of sensor that captures changes in individual pixels rather than generating a complete image frame. This allows them to operate at much higher speeds and with lower power consumption than traditional camera systems. However, this unique characteristic also makes it challenging to develop algorithms for processing the data.
The researchers’ improved dataset consists of high-frequency annotations for pupil detection and tracking, saccade detection, and blink detection. These annotations were generated using a semi-automatic pipeline that combines automated and human-verified steps. The pipeline uses machine learning algorithms to detect eye movements and then corrects any errors made by the algorithms through manual verification.
The resulting dataset contains over 114,000 annotated frames, with each frame capturing the movement of the pupil at 200 Hz. This is a significant improvement over previous datasets, which often had limited annotations or were not specifically designed for event-based eye tracking.
The researchers’ work has important implications for the development of real-time eye-tracking systems that can operate in low-power environments, such as smart glasses and contact lenses. These systems could have applications in various fields, including augmented reality, virtual reality, gaming, and even medical research.
One potential application is in the field of human-computer interaction. With high-frequency eye tracking, users could control devices with their eyes, allowing for more intuitive and efficient interactions. This could be particularly useful for people with disabilities who may struggle to use traditional input methods.
Another potential application is in medical research. High-frequency eye tracking could help researchers better understand various eye movements and diseases, such as Parkinson’s disease or glaucoma. This information could lead to the development of new treatments and therapies.
The researchers’ work has also opened up new avenues for exploring the capabilities of event cameras.
Cite this article: “High-Frequency Eye Tracking Advances with Improved Dataset for Event-Based Eye Movement Detection”, The Science Archive, 2025.
Eye Tracking, Event-Based, Augmented Reality, Virtual Reality, Machine Learning, Computer Vision, Smart Eyewear, Pupil Detection, Saccade Detection, Blink Detection
