Large Scale Image Compression for Benchmarking Supervised Quantum Machine Learning Models

Preparing quantum states for classical data in quantum machine learning (QML) is typically exponentially costly, which poses a significant bottleneck to implementations of QML algorithms. To address these scaling issues, we propose an efficient algorithm for encoding compressed images that is applicable to both grayscale and color images. We demonstrate the effectiveness of our algorithm by encoding standard machine learning datasets, including FashionMNIST, CIFAR10, and Imagenette. In the second stage, we benchmark supervised quantum machine learning models against classical classifiers using these standardized datasets. Our results offer insights into the scalability of QML models for classical data.