SNN / DVS Projects
High-speed classification of event-camera streams using sparse, asynchronous representations optimised for low-latency neuromorphic pipelines.
Event Stream / Classification Animation placeholder
Replace with a DVS event-stream visualisation or classification demo video.
Dynamic Vision Sensors (DVS) output asynchronous streams of events rather than frames, enabling ultra-low-latency perception. This project investigates classification methods that exploit the sparse, asynchronous nature of event data to achieve fast, energy-efficient recognition without waiting for a full frame accumulation window.
Placeholder — describe the event representation used (e.g., time surfaces, voxel grids, spike trains, raw event lists) and the rationale for the chosen approach in the context of low-latency classification.
Placeholder — describe the network architecture, whether SNN, graph-based, or hybrid, and how it processes events asynchronously or in micro-batches to minimise classification latency.
Placeholder for accuracy, latency (time-to-first-classification), and energy metrics. Add benchmark comparisons against frame-based baselines here.