Event Transformer

TL;DR

This paper introduces Event Transformer, a novel event-based representation and attention mechanism that preserves detailed spatiotemporal information for vision tasks, achieving competitive results with minimal computational resources.

Contribution

It proposes a token-based event representation and a Three-way Attention mechanism, enhancing spatiotemporal correlation modeling in event-based vision tasks.

Findings

Competitive performance on object classification and optical flow estimation

Minimal computational resource requirements

Effective preservation of event sequence details

Abstract

The event camera's low power consumption and ability to capture microsecond brightness changes make it attractive for various computer vision tasks. Existing event representation methods typically convert events into frames, voxel grids, or spikes for deep neural networks (DNNs). However, these approaches often sacrifice temporal granularity or require specialized devices for processing. This work introduces a novel token-based event representation, where each event is considered a fundamental processing unit termed an event-token. This approach preserves the sequence's intricate spatiotemporal attributes at the event level. Moreover, we propose a Three-way Attention mechanism in the Event Transformer Block (ETB) to collaboratively construct temporal and spatial correlations between events. We compare our proposed token-based event representation extensively with other prevalent methods for object classification and optical flow estimation. The experimental results showcase its competitive performance while demanding minimal computational resources on standard devices. Our code is publicly accessible at \url{https://github.com/NJUVISION/EventTransformer}.