---
title: JAAD
canonical_url: "https://www.modelscope.cn/datasets/OmniData/JAAD"
md_url: "https://www.modelscope.cn/datasets/OmniData/JAAD.md"
repository: OmniData/JAAD
last_updated: 2024-07-10
license: "[MIT]"
storage_size: "2.9 GB"
domain:
  - publishDate
  - publishUrl
  - paperUrl
  - displayName
  - publisher
  - taskTypes
tasks:
  - 2017
  - "http://data.nvision2.eecs.yorku.ca/JAAD_dataset/"
  - "https://openaccess.thecvf.com/content_ICCV_2017_workshops/papers/w3/Rasouli_Are_They_Going_ICCV_2017_paper.pdf"
  - JAAD
  - "New York University"
  - "Object Detection"
downloads: 65
stars: 1
---

# JAAD

> JAAD - OmniData 在 ModelScope 开源的数据集。displayName: JAAD license: MIT paperUrl: https://openaccess.thecvf.com/contentICCV2017workshops/papers/w3/RasouliAreTheyGoingICCV2017paper.pdf publishDate: "2017" publishUrl: http://data.nvision2.eecs.yorku.ca/JAADdataset/…

OmniData/JAAD 是 ModelScope 魔搭社区上的2017、http://data.nvision2.eecs.yorku.ca/JAAD_dataset/、https://openaccess.thecvf.com/content_ICCV_2017_workshops/papers/w3/Rasouli_Are_They_Going_ICCV_2017_paper.pdf数据集，涉及 publishDate、publishUrl、paperUrl 领域，存储大小 2.9 GB，采用 [MIT] 许可。

- **Repository**: OmniData/JAAD
- **License**: [MIT]
- **Tasks**: 2017, http://data.nvision2.eecs.yorku.ca/JAAD_dataset/, https://openaccess.thecvf.com/content_ICCV_2017_workshops/papers/w3/Rasouli_Are_They_Going_ICCV_2017_paper.pdf, JAAD, New York University, Object Detection
- **Domain**: publishDate, publishUrl, paperUrl, displayName, publisher, taskTypes
- **Storage size**: 2.9 GB
- **Downloads**: 65
- **Stars**: 1
- **Last updated**: 2024-07-10

Source: https://www.modelscope.cn/datasets/OmniData/JAAD

---

displayName: JAAD
license:
- MIT
paperUrl: https://openaccess.thecvf.com/content_ICCV_2017_workshops/papers/w3/Rasouli_Are_They_Going_ICCV_2017_paper.pdf
publishDate: "2017"
publishUrl: http://data.nvision2.eecs.yorku.ca/JAAD_dataset/
publisher:
- New York University
tags:
- Pedestrians
- Traffic lights
taskTypes:
- Object Detection

---
# 数据集介绍
  ## 简介
  JAAD是在自动驾驶的背景下研究共同注意力的数据集。重点是过马路时的行人和驾驶员行为以及影响他们的因素。为此，JAAD数据集提供了从超过240小时的驾驶镜头中提取的346短视频剪辑 (5-10秒长) 的丰富注释集合。这些在北美和东欧的几个地点拍摄的视频代表了在各种天气条件下日常城市驾驶的典型场景。为所有行人提供了带有遮挡标签的边界框，使该数据集适合行人检测。行为注释为与驾驶员互动或需要驾驶员注意的行人指定了行为。对于每个视频，都有几个标签 (天气，位置等) 和固定列表中的带时间戳的行为标签 (例如，停止，行走，观看等)。此外，为每个行人提供人口统计属性列表 (例如年龄、性别、运动方向等) 以及每个帧的可见交通场景元素列表 (例如停车标志、交通信号等)。
  
## Download dataset
:modelscope-code[]{type="git"}