search

Zhongdao / Towards-Realtime-MOT

Joint Detection and Embedding for fast multi-object tracking
MIT License
2.37k stars 539 forks source link

I have two questions about creating a custom datasets. #163

Closed yoonsoo-Shin closed 4 years ago

yoonsoo-Shin commented 4 years ago

Thank you for your great research. I have two questions about creating a custom datasets.

  1. Do you have tools for annotation and labeling in new videos(images) that can be applied to your code? My annotation tool cannot enter [identity].
  2. Does [identity] mean the same person in a video? If 40 people appear in one video, is it correct that 0 to 39 are used for [identity]?

I'm eager to train with my own data.

yoonsoo-Shin commented 4 years ago
  1. I used a similar tool to extract [x, y, w, h, label] and changed the format of the text file.
  2. As a result of reading other Issues, I found out that I was right.
mheriyanto commented 3 years ago 
1. I used a similar tool to extract [x, y, w, h, label] and changed the format of the text file.

2. As a result of reading other Issues, I found out that I was right.

Thank you @yoonsoo-Shin . I have got insight from you.