Electromagnetic fields scattered via PEC MNIST objects

A synthetic modification of the MNIST dataset for EM/RF processing. Each MNIST digit is considered to be a Perfect Electric Conductor (PEC) object. Fixed-position transmitter(s) send monochromatic EM fields (narrowband beacon signals) toward them, and the reflected/diffracted signals/EM fields are collected by fixed-location receivers. The dataset is useful for machine learning applications such as RF imaging and RF target classification. The EM model is derived fromt he paper:K. Stylianopoulos, P. Gavriilidis, G. Gradoni and G. C. Alexandropoulos, "Graph-CNNs for RF Imaging: Learning the Electric Field Integral Equations," 2025 33rd European Signal Processing Conference (EUSIPCO), Palermo, Italy, 2025, pp. 905-909, doi: 10.23919/EUSIPCO63237.2025.11226552.The received signals can be found in the file "E_fields.npy", the corresponding original MNIST images are inside the "DoI_images.npy" and their corresponding labels in "mnist_labels.npy".