Scholarlink

<p>Event cameras like Dynamic Vision Sensors (DVS) report micro-timed brightness changes instead of full frames, offering low latency, high dynamic range, and motion robustness. <b>DVS-PedX</b> (Dynamic Vision Sensor Pedestrian eXploration) is a neuromorphic dataset designed for pedestrian detection and crossing-intention analysis in normal and adverse weather conditions across two complementary sources: (1) synthetic event streams generated in the CARLA simulator for controlled “approach–cross” scenes under varied weather and lighting, comprising 198 sequences (117 good weather, 81 bad weather) with 178,200 total frames; and (2) real-world JAAD dash-cam videos (346 clips) converted to event streams using the <i>v2e</i> tool, preserving natural behaviors and backgrounds. Each sequence includes paired RGB frames, per-frame DVS “event frames” (33 ms accumulations), and frame-level binary labels (crossing vs. not crossing). We provide raw AEDAT 2.0/AEDAT 4.0 event files, AVI DVS video files, and metadata for flexible re-processing. Baseline experiments using spiking neural networks (SNNs) with SpikingJelly achieve an F1-score of 86.37% on the synthetic validation set, revealing a sim-to-real gap that motivates domain adaptation and multimodal fusion. DVS-PedX aims to accelerate research in event-based pedestrian safety, intention prediction, and neuromorphic perception.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

DVS-PedX: Synthetic-and-Real Event-Based Pedestrian Dataset

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月6日

查看全文
  • Mustafa Sakhai,
  • Kaung Sithu,
  • Min Khant Soe Oke,
  • Maciej Wielgosz

摘要

Event cameras like Dynamic Vision Sensors (DVS) report micro-timed brightness changes instead of full frames, offering low latency, high dynamic range, and motion robustness. DVS-PedX (Dynamic Vision Sensor Pedestrian eXploration) is a neuromorphic dataset designed for pedestrian detection and crossing-intention analysis in normal and adverse weather conditions across two complementary sources: (1) synthetic event streams generated in the CARLA simulator for controlled “approach–cross” scenes under varied weather and lighting, comprising 198 sequences (117 good weather, 81 bad weather) with 178,200 total frames; and (2) real-world JAAD dash-cam videos (346 clips) converted to event streams using the v2e tool, preserving natural behaviors and backgrounds. Each sequence includes paired RGB frames, per-frame DVS “event frames” (33 ms accumulations), and frame-level binary labels (crossing vs. not crossing). We provide raw AEDAT 2.0/AEDAT 4.0 event files, AVI DVS video files, and metadata for flexible re-processing. Baseline experiments using spiking neural networks (SNNs) with SpikingJelly achieve an F1-score of 86.37% on the synthetic validation set, revealing a sim-to-real gap that motivates domain adaptation and multimodal fusion. DVS-PedX aims to accelerate research in event-based pedestrian safety, intention prediction, and neuromorphic perception.