Event-based Spiking Neural Networks for Object Detection: A Review of Datasets, Architectures, Learning Rules, and Implementation

TL;DR

This paper presents a systematic review of datasets, architectures, learning methods, implementation techniques, and evaluation methodologies used in CV-based object detection tasks using SNNs, and codifies the effectiveness of fully connected, convolutional, and recurrent architectures.

Abstract

Spiking Neural Networks (SNNs) represent a biologically inspired paradigm offering an energy-efficient alternative to conventional artificial neural networks (ANNs) for Computer Vision (CV) applications. This paper presents a systematic review of datasets, architectures, learning methods, implementation techniques, and evaluation methodologies used in CV-based object detection tasks using SNNs. Based on an analysis of 151 journal and conference articles, the review codifies: 1) the effectiveness of fully connected, convolutional, and recurrent architectures; 2) the performance of direct unsupervised, direct supervised, and indirect learning methods; and 3) the trade-offs in energy consumption, latency, and memory in neuromorphic hardware implementations. An open-source repository along with detailed examples of Python code and resources for building SNN models, event-based data processing, and SNN simulations are provided. Key challenges in SNN training, hardware integration, and future directions for CV applications are also identified.

Figures (14)

• Figure 1: Overview of the PRISMA approach used for conducting the literature review for this study.
• Figure 2: (Left) Different architectures discussed in the reviewed papers. The most discussed architecture is SCNNs. (Right) Distribution of dataset types reported in the reviewed articles.
• Figure 3: (Left) The distribution of learning methods employed in the reviewed papers. (Right) The distribution of implementation mediums for SNNs across the reviewed papers, distinguishing between software tools and hardware-based implementations.
• Figure 4: The systematic review presented in this paper encompasses publications from 2000 to 2023. Starting around 2015, there was a noticeable uptick in activity. This increase coincides with the growing traction of SNNs within the research community.
• Figure 6: A biological neuron with its main components: soma, dendrites, axon, and axon terminals. Dendrites receive signals, which are integrated in the soma. The inset depicts neurotransmitter release at the synaptic cleft, enabling communication with other neurons.