A Non‐Volatile Electrochromic Organic Electrochemical Transistor for Visual Neural Networks Featuring Color‐Encoded Thresholds

The integration of neuromorphic computing with optical display technology into a unified device is essential for advancing intelligent vision systems. However, existing devices are constrained by inherent optoelectronic inconsistencies and insufficient non‐volatile characteristics, which typically result in single‐threshold activation rather than the preferred programmable multi‐threshold processing. This limitation leads to suboptimal performance in complex visual tasks. This study introduces a two‐in‐one non‐volatile electrochromic organic electrochemical transistor (NCOECT), in which the semiconducting polymer poly(2‐(3,3′‐bis(2‐(2‐(2‐methoxyethoxy)ethoxy)ethoxy)‐[2,2′‐bithiophen]‐5‐yl)thieno[3,2‐b]thiophene) (P(g2T‐TT)) simultaneously serves as both the switching channel and the chromatic medium. This configuration establishes a precise and enduring correlation between electrical conductance and visible color. The best‐performing device exhibits intrinsic non‐volatility, maintaining its state for over 10 000 s, and offers 120 distinct conductance levels. An implemented visual artificial neural network (ANN) based on the device achieves handwritten digit recognition with 96.1% accuracy. Notably, by leveraging the one‐to‐one correspondence between conductance and color, the study introduces color encoding as a programmable threshold mechanism. Using visible color states as physical filters significantly enhances classification performance on noisy images. This work offers an integrated device solution that combines filtering, computation, and display functionalities within a single platform, enabling reliable, adaptive, and visually interpretable image recognition under noisy conditions.