---
title: Caltech_Pedestrian_Detection_etc
canonical_url: "https://www.modelscope.cn/datasets/OmniData/Caltech_Pedestrian_Detection_etc"
md_url: "https://www.modelscope.cn/datasets/OmniData/Caltech_Pedestrian_Detection_etc.md"
repository: OmniData/Caltech_Pedestrian_Detection_etc
last_updated: 2024-07-12
license: "Apache License 2.0"
storage_size: "23 GB"
downloads: 1019
stars: 0
---

# Caltech_Pedestrian_Detection_etc

> Caltech_Pedestrian_Detection_etc - OmniData 在 ModelScope 开源的数据集。displayName: Caltech Pedestrian Detection Benchmark labelTypes: Box2D mediaTypes: Video paperUrl:…

OmniData/Caltech_Pedestrian_Detection_etc 是 ModelScope 魔搭社区上的数据集，存储大小 23 GB，采用 Apache License 2.0 许可。

- **Repository**: OmniData/Caltech_Pedestrian_Detection_etc
- **License**: Apache License 2.0
- **Storage size**: 23 GB
- **Downloads**: 1019
- **Stars**: 0
- **Last updated**: 2024-07-12

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

---

displayName: Caltech Pedestrian Detection Benchmark
labelTypes:
- Box2D
mediaTypes:
- Video
paperUrl: "http://www.vision.caltech.edu/Image_Datasets/CaltechPedestrians/files/CVPR09pedestrians.pdf\r\nhttp://www.vision.caltech.edu/Image_Datasets/CaltechPedestrians/files/PAMI12pedestrians.pdf"
publishDate: "2009"
publishUrl: https://computervisiononline.com/dataset/1105138627
publisher:
- California Institute of Technology
tags:
- Pedestrian
taskTypes:
- Object Tracking

---
# 数据集介绍
  ## 简介
  加州理工学院行人数据集由大约 10 小时的 640x480 30Hz 视频组成，该视频取自在城市环境中通过常规交通行驶的车辆。注释了大约 250,000 帧（在 137 个大约分钟长的片段中），总共 350,000 个边界框和 2300 个独特的行人。注释包括边界框和详细的遮挡标签之间的时间对应关系。更多信息可以在我们的 PAMI 2012 和 CVPR 2009 基准测试文件中找到。
  ## 类定义
  ```
1:people
2:person
3:person?
4:person-fa
```
  ## 引文
  ```
@article{dollar2011pedestrian,
  title={Pedestrian detection: An evaluation of the state of the art},
  author={Dollar, Piotr and Wojek, Christian and Schiele, Bernt and Perona, Pietro},
  journal={IEEE transactions on pattern analysis and machine intelligence},
  volume={34},
  number={4},
  pages={743--761},
  year={2011},
  publisher={IEEE}
}
```
  
## Download dataset
:modelscope-code[]{type="git"}