行车记录仪数据集_行人检测数据集(极市平台的补充版)

“极市平台” 已经提供了一些数据集到百度网盘:极市平台：行人检测数据集汇总(持续更新)​zhuanlan.zhihu.com

下面这些数据集是“极市平台”尚未收录的：

具体描述如下：

获取方式：账号注册需审核，注册通过即可下载，共103GB左右。

1.ECP_day_img_train.zip 50.2GB

ECP training images (day) from 31 European cities to be used to train your algorithm.

2.ECP_day_img_val.zip 9.0GB

ECP validation images (day) from 31 European cities to be used to validate your algorithm.

3.ECP_day_img_test.zip 25.3GB

ECP test images (day) from 31 European cities to be used to test your algorithm. Labeles for these images are not provided. Test results can be submitted to allow a fair comparison to other algorithms.

4.ECP_day_labels_train.zip 12.6MB

ECP training annotations for 'Images Training (Day)' to validate your algorithm.

5.ECP_day_labels_val.zip 2.3MB

ECP validation annotations for 'Images Validation (Day)' to validate your algorithm.

6.ECP_night_img_train.zip 9.4GB

ECP training images (night) from 31 European cities to be used to train your algorithm.

7.ECP_night_img_val.zip 1.7GB

ECP validation images (night) from 31 European cities to be used to validate your algorithm.

8.ECP_night_img_test.zip 4.7GB

ECP test images (night) from 31 European cities to be used to test your algorithm. Labeles for these images are not provided. Test results can be submitted to allow a fair comparison to other algorithms.

9.ECP_night_labels_train.zip 2.2MB

ECP training annotations for 'Images Training (Night)' to validate your algorithm.

10.ECP_night_labels_val.zip 410KB

ECP training annotations for 'Images Validation (Night)' to validate your algorithm.

以下介绍摘自论文：The images for this dataset were collected on-board a moving vehicle in 31 cities of 12 European countries. With over 238200 person instances manually labeled in over 47300 images, EuroCity Persons is nearly one order of magnitude larger than person datasets used previously for benchmarking. The dataset furt