AI-powered retinal projection for immersive autostereoscopic display

Autostereoscopic displays promise immersive visual experiences without the need for glasses while delivering parallax images to the appropriate retinas requires high precision. This necessitates dynamically compensating for instant changes in viewing position, gaze, and posture to overcome system latency. Here we report an artificial intelligence (AI) assisted retinal projection framework integrated into a time-sequential directional backlight autostereoscopic display, enabling stable and high-resolution 3D perception for desktop devices. By combining retina resolved imaging with real-time eye motion prediction, the system compensates end-to-end latency and dynamically aligns directional light delivery with the viewer’s retinal position. Experiments using both controlled motion phantoms and human subjects demonstrate substantial reductions in undesirable ghost images and improved stereoscopic stability during viewer movement. Unlike conventional multi view displays with discretized viewpoints, the proposed architecture supports a seamless viewing zone while preserving full panel resolution for each eye. This work establishes latency aware retinal projection as a viable paradigm for practical autostereoscopic displays and provides a scalable foundation for future immersive display technologies.