How to calculate category_embedding for my own dataset?

[ljjyxz123]

In config.py, line 8, there is category_embedding definition:
https://github.com/detectRecog/PointTrack/blob/master/config.py#L8
category_embedding = [[0.9479751586914062, 0.4561353325843811, 0.16707628965377808], [0.1,-0.1,0.1], [0.5455077290534973, -0.6193588972091675, -2.629554510116577], [-0.1,0.1,-0.1]]

How is this value calculated?
If I want to train my own dataset, how to calculate the category_embedding for my own category?

Thanks! @detectRecog