Random-CSG-3D-ECT: Synthetic 3D Electrical Capacitance Tomography Dataset for 32-Electrode Volumetric Reconstruction

This dataset contains synthetic three-dimensional electrical capacitance tomography (3D ECT) data generated for a two-ring 32-electrode sensor configuration. Each sample consists of a capacitance measurement vector and the corresponding 64 × 64 × 64 volumetric relative-permittivity distribution. The data were generated using the open-source ECTsim simulation framework and were used to train and evaluate a machine-learning model for direct capacitance-to-volume reconstruction in 3D ECT. The simulated phantoms were constructed from randomly placed and transformed geometric primitives, including ellipsoids, cylinders, and cuboids. These primitives were combined using constructive solid geometry operations, including union, subtraction, and intersection, to create diverse synthetic 3D object distributions. The dataset is intended to support research on volumetric ECT reconstruction, synthetic-data-trained inverse models, sensor-specific 3D tomography, and real-time-capable measurement-to-image workflows.