Eye tracking is used to record the position of human eyes and their movements, primarily to enable machines to perceive and interact with humans, especially in Augmented and Virtual Reality (AR/VR) settings. For instance, foveated rendering, which relies on high-performance eye tracking, stands as one of the core technologies enabling an immersive user experience in AR/VR devices.

However, despite the desired attributes of high throughput (e.g., >240 FPS), small form factor, and enhanced visual privacy for supporting real-time human-machine interaction on resource-constrained AR/VR devices, existing eye tracking systems utilize bulky lens-based cameras. Consequently, they face challenges related to both large form factors and high communication costs between the camera and backend processor, which limit the achievable throughput. Moreover, the growing demand to integrate high-performance yet complex AI algorithms, such as deep neural networks (DNNs), into AR/VR devices further compounds the difficulty of achieving high throughput.

To address these challenges, we are pioneering camera, algorithm, and processor chip co-designed intelligent eye-tracking systems to meet the requirements without compromising the eye tracking accuracy. Welcome to join us in paving the way for next-generation eye tracking solutions and sheding light on future innovations for intelligent imaging systems.

Corresponding Publications:

• EyeCoD @ IEEE Micro Top Picks'23
• EyeCoD @ ISCA'22
• i-FlatCam @ VLSI'22
• Best Demo @ DAC'22