A brain-inspired paradigm for scalable quantum vision
Abstract
One of the fundamental tasks in machine learning is image classification, which serves as a key benchmark for validating algorithm performance and practical potential. However, effectively processing high-dimensional, detail-rich images, a capability that is inherent in biological vision, remains a persistent challenge. Inspired by the human brain's efficient ``Forest Before Trees'' cognition, we propose a novel Guiding Paradigm for image recognition, leveraging classical neural networks to analyze global low-frequency information and guide targeted quantum circuit towards critical high-frequency image regions. We present the Brain-Inspired Quantum Classifier (BIQC), implementing this paradigm via a complementarity architecture where a quantum pathway analyzes the localized intricate details identified by the classical pathway. Numerical simulations on diverse datasets, including high-resolution images, show the BIQC's superior accuracy and scalability compared to existing methods. This highlights the promise of brain-inspired, hybrid quantum-classical approach for developing next-generation visual systems.
Cite
@article{arxiv.2509.05919,
title = {A brain-inspired paradigm for scalable quantum vision},
author = {Chenghua Duan and Xiuxing Li and Wending Zhao and Lin Yao and Qing Li and Ziyu Li and Fukang Li and Junhao Ma and Xia Wu},
journal= {arXiv preprint arXiv:2509.05919},
year = {2025}
}