Abstract
Event cameras offer unparalleled advantages for real-time perception indynamic environments, thanks to the microsecond-level temporal resolution andasynchronous operation. Existing event detectors, however, are limited byfixed-frequency paradigms and fail to fully exploit the high-temporalresolution and adaptability of event data. To address these limitations, wepropose FlexEvent, a novel framework that enables detection at varyingfrequencies. Our approach consists of two key components: FlexFuse, an adaptiveevent-frame fusion module that integrates high-frequency event data with richsemantic information from RGB frames, and FlexTune, a frequency-adaptivefine-tuning mechanism that generates frequency-adjusted labels to enhance modelgeneralization across varying operational frequencies. This combination allowsour method to detect objects with high accuracy in both fast-moving and staticscenarios, while adapting to dynamic environments. Extensive experiments onlarge-scale event camera datasets demonstrate that our approach surpassesstate-of-the-art methods, achieving significant improvements in both standardand high-frequency settings. Notably, our method maintains robust performancewhen scaling from 20 Hz to 90 Hz and delivers accurate detection up to 180 Hz,proving its effectiveness in extreme conditions. Our framework sets a newbenchmark for event-based object detection and paves the way for moreadaptable, real-time vision systems.