Bio‐Inspired Multimodal InSe Vision Sensor via Laser‐Induced Asymmetric Contacts for Dynamic Trajectory Extraction

ABSTRACT Neuromorphic vision systems demonstrate exceptional capabilities in spatial perception. However, identifying dynamic targets in complex environments often faces challenges from redundant background data, which imposes unnecessary computational overhead. Inspired by the frog retina's ability to suppress static backgrounds while detecting motion at the sensory level, we develop a multimodal vision sensor on a single 2D InSe platform. We fabricate asymmetric Schottky contacts within the device by a facile laser‐induced surface‐modification technique. This creates a strong built‐in electric field, enabling efficient self‐powered photodetection. Under an applied bias, defect‐mediated charge trapping facilitates an opto‐synaptic response. By switching between these two distinct modes, the InSe device functions as reconfigurable convolutional kernels, achieving high‐contrast spatial edge detection. Furthermore, this multimodal capability allows the sensors to isolate dynamic motion trajectories from redundant backgrounds, mimicking the frog's efficient visual processing. This work demonstrates a strategy for highly efficient motion perception in complex scenarios, and paves the way for developing low‐power and multi‐dimensional intelligent sensing systems.