Giters

nkicsl / NKUP-dataset

A benchmark cross clothes re-ID dataset

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Reading Annotations

Ehsan-Yaghoubi opened this issue · comments

commented

Hello, Could you please help me with the following questions?

How to read the annotations?
For example, in 0005_D0912_C03S01_00151.jpg I realized that

  • 0005 is the person ID
  • D0912 is the recording day
  • C03S01 is the camera number (what is S01?)
  • 00151 is the frame number

But steel I cannot separate the images based on the person ID and clothing style. For example, I want the images of ID 1 in Folder 1 and his clothing 1 in subfolder 1 and clothing 2 in subfolder 2. Is it possible?

When I tried to do it based on the day, I saw some ids do not change their clothing on different days. Also, sometimes in a single day, one ID may appear with different clothing. So, How do you solve this issue?

commented

Hello, Could you please help me with the following questions?

How to read the annotations? For example, in 0005_D0912_C03S01_00151.jpg I realized that

  • 0005 is the person ID
  • D0912 is the recording day
  • C03S01 is the camera number (what is S01?)
  • 00151 is the frame number

But steel I cannot separate the images based on the person ID and clothing style. For example, I want the images of ID 1 in Folder 1 and his clothing 1 in subfolder 1 and clothing 2 in subfolder 2. Is it possible?

When I tried to do it based on the day, I saw some ids do not change their clothing on different days. Also, sometimes in a single day, one ID may appear with different clothing. So, How do you solve this issue?

  1. Each camera in NKUP collected multiple continuous videos in a day. CxxS01 means the first video shot by camera xx on the recording day. Similarly, S02 means the second video by camera xx.
  2. Sorry, we don't give the classification of different clothes in NKUP. But, we are currently carrying out the division and expansion of the dataset.