AI-assisted metaphotonics: A Comprehensive Review of Artificial Intelligence-Driven Approaches for Metaphotonic Systems

• • End-to-end metasurface design for temperature imaging via broadband Planck-radiation regression achieved a 13% improvement in temperature estimation accuracy compared to conventional methods, with a root-mean-square error (RMSE) of 0.8 K across a 300–500 K range, enabling non-contact thermal imaging for industrial process monitoring. • • Neural nano-optics for high-quality thin lens imaging demonstrated a 2.5× reduction in chromatic aberration and a 40% increase in modulation transfer function (MTF) at 50 cycles/mm compared to traditional metalenses, directly impacting compact camera modules for mobile devices and AR/VR systems. • • End-to-end optimization of metalens for broadband and wide-angle imaging achieved a 70% average focusing efficiency across 400–700 nm and a 60° field of view with <5% distortion, surpassing the 45% efficiency and 40° FOV of conventional designs, critical for automotive LiDAR and machine vision. • • Deep-learning-based colorimetric polarization-angle detection with metasurfaces attained a polarization angle resolution of 0.1° and a detection accuracy of 98.5% under ambient light, outperforming commercial polarimeters by 3× in speed and 2× in compactness, with direct implications for remote sensing and biomedical diagnostics.