A 25-mus/inf Event-driven Graph Neural Network Processor with Spatiotemporal Caching and Spline Convolution for Ultra-low-latency AI at the Edge

arXiv:2609.15241v1 Announce Type: new Abstract: Dynamic-vision-sensor (DVS) cameras generate events on a per-pixel basis with a mus-level temporal resolution, calling for new algorithm-hardware co-design approaches compared to standard frame-based vision. While event-driven graph neural networks (EV-GNNs) emerge as a promising algorithmic solution, they raise new HW challenges by mixing dense-regular compute operations and sparse-irregular memory accesses. We present ETHEREAL, the first EV-GNN accelerator that scales to 640imes480 resolutions, thanks to a neighbor-parallel spline convolution engine and a 2D/3D-split memory hierarchy with a novel region-of-interest spatiotemporal caching mechanism. Measurement results demonstrate end-to-end inference with 25.6mus latency and 1.7muJ energy per event on state-of-the-art workloads

ETHEREAL: A 25.6-mus/inf. Low-latency Event-driven Graph-neural-network Processor for High-resolution Vision at the Edge
Hardware-aware Graph Neural Networks prunning for embedded event-based vision
Rethinking Event-Based Object Dtection through Representation-Level Temporal Aggregation and Model-Level Hypergraph Reasoning

Source: arXiv cs.CV | 2026-09-15

Related