FRED: The Florence RGB-Event Drone Dataset
收藏资源简介:
FRED数据集是一个用于无人机检测、跟踪和轨迹预测的多模态数据集,结合了RGB视频和事件流。数据集包含超过7小时的密集标注无人机轨迹,使用了5种不同的无人机模型,并包括雨和不良光照条件等具有挑战性的场景。该数据集旨在解决传统RGB相机在捕捉快速移动对象时的局限性,特别是对于小型、快速、轻量级的无人机。FRED数据集为无人机感知和多功能时空理解的研究提供了新的基准和资源。
The FRED dataset is a multimodal dataset for unmanned aerial vehicle (UAV) detection, tracking and trajectory prediction, which integrates RGB videos and event streams. It contains over 7 hours of densely annotated UAV trajectories, utilizes 5 distinct UAV models, and encompasses challenging scenarios including rainy weather and poor lighting conditions. This dataset is designed to address the limitations of conventional RGB cameras when capturing fast-moving objects, particularly small, fast and lightweight UAVs. The FRED dataset offers a novel benchmark and resource for research on UAV perception and versatile spatiotemporal understanding.
FRED: The Florence RGB-Event Drone Dataset
作者信息
- Gabriele Magrini (MICC, University of Florence, Italy)
- Niccolò Marini (MICC, University of Florence, Italy)
- Federico Becattini (University of Siena, Italy)
- Lorenzo Berlincioni (MICC, University of Florence, Italy)
- Niccolò Biondi (MICC, University of Florence, Italy)
- Pietro Pala (MICC, University of Florence, Italy)
- Alberto Del Bimbo (MICC, University of Florence, Italy)
数据集概述
- 名称:Florence RGB-Event Drone Dataset (FRED)
- 类型:多模态数据集(RGB视频 + 事件流)
- 用途:无人机检测、跟踪和轨迹预测
- 特点:
- 超过7小时的密集标注无人机轨迹
- 包含5种不同无人机模型
- 涵盖挑战性场景(如雨天和不利光照条件)
数据集优势
- 针对高速运动物体捕获的局限性(传统RGB相机)
- 事件相机提供高时间分辨率和动态范围
- 填补现有基准测试在精细时间分辨率和无人机特定运动模式方面的不足
评估信息
- 提供详细的评估协议
- 包含每项任务的标准指标
- 支持可复现的基准测试
相关链接
- 论文代码:https://miccunifi.github.io/FRED/
- arXiv:https://miccunifi.github.io/FRED/
- 数据集:https://miccunifi.github.io/FRED/

- 1FRED: The Florence RGB-Event Drone Dataset佛罗伦萨大学 · 2025年



