Event-LSTM: An Unsupervised and Asynchronous Learning-Based Representation for Event-Based Data

Event cameras are activity-driven bio-inspired vision sensors that respond asynchronously to intensity changes resulting in sparse data known as events. It has potential advantages over conventional cameras, such high temporal resolution, low latency, and power consumption. Given the asynchronous spatio-temporal nature of data, event processing is predominantly solved by transforming events into a $2D$ spatial grid representation applying standard pipelines. In this work, we propose an auto-encoder architecture named Event-LSTM generate representation. Ours following main 1) Unsupervised, task-agnostic learning grid. ideally suited for domain, where task-specific labeled scarce, 2) Asynchronous sampling This leads speed invariant energy-efficient Evaluations on appearance-based motion-based tasks demonstrate our approach yields improvement state-of-the-art techniques while providing flexibility learn from unlabelled data.