Spiking Neural Networks in Imaging: A Review and Case Study

This review examines the state of spiking neural networks (SNNs) for imaging, combining a structured literature survey, a comparative meta-analysis of reported datasets, training strategies, hardware platforms, and applications and a case study on LMU-based depth estimation in direct Time-of-Flight (dToF) imaging. While SNNs demonstrate promise for energy-efficient, event-driven computation, current progress is constrained by reliance on small or custom datasets, ANN-SNN conversion inefficiencies, simulation-based hardware evaluation, and a narrow focus on classification tasks. The analysis highlights scaling trade-offs between accuracy and efficiency, persistent latency bottlenecks, and limited sensor-hardware integration. These findings were synthesised into key challenges and future directions, emphasising benchmarks, hardware-aware training, ecosystem development, and broader application domains.