Person Re-identification数据集描述——DukeMTMC-reID

dataset

数据集简介

  DukeMTMC 数据集是一个大规模标记的多目标多摄像机行人跟踪数据集。它提供了一个由 8 个同步摄像机记录的新型大型高清视频数据集,具有 7,000 多个单摄像机轨迹和超过 2,700 多个独立人物,DukeMTMC-reID 是 DukeMTMC 数据集的行人重识别子集,并且提供了人工标注的bounding box。

目录结构

DukeMTMC-reID
  ├── bounding_box_test
       ├── 0002_c1_f0044158.jpg
       ├── 3761_c6_f0183709.jpg
       ├── 7139_c2_f0160815.jpg
  ├── bounding_box_train
       ├── 0001_c2_f0046182.jpg
       ├── 0008_c3_f0026318.jpg
       ├── 7140_c4_f0175988.jpg
  ├── query
       ├── 0005_c2_f0046985.jpg
       ├── 0023_c4_f0031504.jpg
       ├── 7139_c2_f0160575.jpg
  └── CITATION_DukeMTMC.txt
  └── CITATION_DukeMTMC-reID.txt
  └── LICENSE_DukeMTMC.txt
  └── LICENSE_DukeMTMC-reID.txt
  └── README.md

目录介绍

从视频中每 120 帧采样一张图像,得到了 36,411 张图像。一共有 1,404 个人出现在大于两个摄像头下,有 408 个人 (distractor ID) 只出现在一个摄像头下。
1) “bounding_box_test”——用于测试集的 702 人,包含 17,661 张图像(随机采样,702 ID + 408 distractor ID)
2) “bounding_box_train”——用于训练集的 702 人,包含 16,522 张图像(随机采样)
3) “query”——为测试集中的 702 人在每个摄像头中随机选择一张图像作为 query,共有 2,228 张图像

命名规则

以 0001_c2_f0046182.jpg 为例
1) 0001 表示每个人的标签编号;
2) c2 表示来自第二个摄像头(camera2),共有 8 个摄像头;
3) f0046182 表示来自第二个摄像头的第 46182 帧。

Dataset Insights

数据分布

Figure. The image distribution of DukeMTMC-reID training set. We note that the median of images per ID is 20. But some ID may contain lots of images, which may compromise some algorithms. (For example, ID 5388 contains 426 images.)

Thank Xun for suggestions.

地理位置

This picture is from DukeMTMC Homepage.

测试协议

(Matlab)To evaluate, you need to calculate your gallery and query feature (i.e., 17661x2048 and 2228x2048 matrix) and save them in advance. Then download the codes in this repository. You just need to change the image path and the feature path in the evaluation_res_duke_fast.m and run it to evaluate.

(Python)We also provide an evaluation code in python. You may refer to here.

下载地址

  1. Google Drive
  2. Baidu Disk 密码:bhbh
  3. DukeMTMC Project

Baseline

We release our baseline training code and pretrained model in [Matconvnet Version] and [Pytorch Version]. You can choose one of the two tools to conduct the experiment. Furthermore, you may try our new Pedestrain Alignment Code which combines person alignment with re-ID.

Or you can directly download the finetuned ResNet-50 baseline feature. You can download it from GoogleDriver or BaiduYun, which includes the feature of training set, query set and gallery set. The DukeMTMC-reID LICENSE is also included.

State-of-the-art

Citation

If you use this dataset, please kindly cite the following two papers:

1
2
3
4
5
6
7
8
9
10
11
12
@inproceedings{zheng2017unlabeled,
title={Unlabeled Samples Generated by GAN Improve the Person Re-identification Baseline in vitro},
author={Zheng, Zhedong and Zheng, Liang and Yang, Yi},
booktitle={Proceedings of the IEEE International Conference on Computer Vision},
year={2017}
}
@inproceedings{ristani2016MTMC,
title = {Performance Measures and a Data Set for Multi-Target, Multi-Camera Tracking},
author = {Ristani, Ergys and Solera, Francesco and Zou, Roger and Cucchiara, Rita and Tomasi, Carlo},
booktitle = {European Conference on Computer Vision workshop on Benchmarking Multi-Target Tracking},
year = {2016}
}

参考文献