ASPIR: A Single-Pixel Imaging Research Platform

ASPIR is an open-source desktop application that unifies the complete single-pixel imaging computational pipeline behind a graphical user interface. It covers dataset generation, mask-pattern design, reconstruction with classical and iterative algorithms, post-processing with ten neural network architectures, and analysis of reconstruction quality, inference latency, and energy consumption. Built-in hardware-backend auto-detection enables cross-platform benchmarking across desktop NVIDIA GPUs, Intel CPUs (RAPL), and NVIDIA Jetson embedded devices. ASPIR is designed to democratise single-pixel imaging experimentation for non-programming researchers and to serve as a pedagogical tool for computational imaging, compressive sensing, and energy-aware deep learning.