LiteEvent-AE: Lightweight Autoencoder for Event-Based Vision on Low-Latency Energy-Constrained Edge Devices

2026-08-22

AI Summary

本文提出了一种轻量级事件驱动自编码器LiteEvent-AE，通过高效的神经形态数据压缩和低计算成本，解决了低延迟、能量受限边缘设备上的事件视觉处理问题。

Abstract

Event-based vision has emerged as a promising paradigm for energy-aware artificial intelligence (AI), offering sparse, low-latency visual signals that reduce redundant data processing and support sustainable edge computing. However, the asynchronous and noise-prone nature of event streams creates challenges for conventional deep learning models, which are often too computationally intensive for low-power embedded platforms. This work presents a compact and configurable event-driven autoencoder that efficiently compresses neuromorphic data while preserving essential spatiotemporal structure for downstream inference. The architecture integrates lightweight convolutional encoding with robust performance under adaptive event thresholding and a minimal classifier head, enabling substantial reductions in computational cost without degrading recognition fidelity. Extensive evaluations on the Smart Event Face Dataset (SEFD) and Event-Based Crossing Dataset (EBCD) show that the proposed framework achieves competitive or superior accuracy compared to YOLOv9 while requiring up to 35.6$\times$ fewer parameters. To assess real-world sustainability, the model is deployed on resource-constrained hardware: a Raspberry Pi 4B and a NVIDIA Jetson Nano. On NVIDIA Jetson Nano, it delivers real-time throughput of 44.8 FPS. On a Raspberry Pi 4B CPU, the 50\% autoencoder classifier consumes 16.19 J for the evaluated inference workload, corresponding to approximately 726.3$\times$ lower energy consumption than YOLOv9 under the same evaluation protocol. These results demonstrate the potential of compact event-driven models to advance environmentally conscious, low-power AI systems for high-speed perception in autonomous, mobile, and embedded computing environments.

Problem

• Event-based Vision
• Low-latency
• Energy-constrained
• Edge Devices
• Deep Learning Models

Innovation

• event-driven autoencoder
• low-latency
• energy-constrained
• lightweight convolutional encoding
• adaptive event thresholding