Event cameras output massive, low-information data streams that overwhelm current models. We propose compressing these into informative "neural events" using discrete codes that trigger only upon state changes. This halves the data rate while matching or exceeding state-of-the-art performance in object detection and classification. - View it on GitHub