Dark-traffic
收藏资源简介:
Dark-traffic是一个专为低光交通场景设计的大规模数据集,包含了超过10,000张图像,共有100,000个标注实例。该数据集支持目标检测、实例分割和光流估计,并使用了一种基于现实世界低光分布的物理光照衰减方法生成。它旨在解决低光环境下目标感知的挑战,为相关研究提供了宝贵的数据资源。
Dark-traffic is a large-scale dataset specifically designed for low-light traffic scenarios, containing over 10,000 images and a total of 100,000 annotated instances. This dataset supports tasks including object detection, instance segmentation and optical flow estimation, and is generated via a physical lighting attenuation method based on real-world low-light distributions. It aims to address the challenges of object perception in low-light environments, providing a valuable data resource for related research.
数据集概述
数据集名称
A Biologically Inspired Separable Learning Vision Model for Real-time Traffic Object Perception in Dark
数据集描述
专为低光照交通条件下的目标感知任务设计的基准数据集,包含目标检测、实例分割和光流估计三个任务。
数据规模
- 图像数量:约10,000张
- 标注数量:约100,000个
数据内容
- 图像类型:低光照交通场景图像
- 标注类型:目标检测、实例分割、光流估计标注
数据获取
数据集可通过Google Drive下载:https://drive.google.com/drive/folders/1B8EzDn64bGBgyRCfppL_jhcOA3hIwnzi?usp=sharing
发布状态
- 图像数据:已发布
- 标注数据:已于9月05日发布(检测/分割/光流标注)
- 相关代码:SLVM静态和运动感知代码准备中(9月07日更新)
相关研究
基于Expert Systems with Applications期刊论文(DOI:10.1016/j.eswa.2025.129529)构建的数据集
技术依赖
- slim-neck-by-gsconv:https://github.com/AlanLi1997/slim-neck-by-gsconv
- rethinking-fpn:https://github.com/AlanLi1997/rethinking-fpn
- ultralytics:https://github.com/ultralytics/ultralytics
- gmflow:https://github.com/haofeixu/gmflow
- NeuFlow_v2:https://github.com/neufieldrobotics/NeuFlow_v2




