Neuromorphic vision systems process information directly at the sensor front end, offering a promising hardware pathway for next-generation machine vision. However, current implementations are typically restricted to single functions and rely on device mechanisms that limit scalability, uniformity and miniaturization, highlighting the need for multifunctional and integrable neuromorphic hardware. Herein, we introduce a photo-regulated doping strategy enabled by galvinoxyl radicals (GX) as wavelength-dependent bipolar dopants, in which radical-mediated charge transfer enables wavelength-dependent optical switching between excitatory and inhibitory responses within a single-layer neuromorphic device. This mechanism supports high photoresponse and programmable synaptic dynamics, allowing the construction of a large‑scale (256 × 256) neuromorphic vision system that integrates dynamic motion detection, spatiotemporal memory, and contrast-enhanced edge extraction. This work provides a potent hardware solution for complex dynamic scenarios, establishing a scalable pathway toward fully integrated, high-performance neuromorphic vision system.
Qiu et al. (Wed,) studied this question.