Zhenghao Long, Kaichen Wang, Shaoshuai He, Xiaofei Sun, Zilong Ye, Jianye Li, Ke Shen, Xiao Qiu, Yucheng Ding, Feng Xue, Yu Zhou, Shivam Kumar, Swapnadeep Poddar, Xichao Tan, Yunlong Zi, Zhiyong Fan
Vision sensors are central to machine vision and intelligence. However, conventional planar devices rely on multielement optics and external processors to correct optical aberrations. This approach constrains miniaturization, reduces power efficiency and generally limits the field of view. Curved sensors can provide compact, aberration-corrected imaging, but their resolution has been below practicable levels. Here we present a bioinspired bionic eye system based on a hemispherical tandem artificial retina. This high curvature image sensor achieves a pixel density of 1,905 ppi with 367,500 total pixels, providing full-colour imaging across 300-800 nm and an aberration-corrected field of view exceeding 160°. Furthermore, the tandem design enables in-sensor, event-driven motion detection. Compared with frame-based imaging, this approach reduces the demand for bandwidth by over 99.95% and achieves a motion recognition accuracy of 98.6%. This work addresses the resolution bottleneck of hemispherical sensors and highlights their promise as compact, multifunctional vision applications.