Eager to know the training process?

[SzuPc]

I know the path of thinking, the final result of the tracker is to set the number of each individual, is it the case that the number of the main body has progressed as a model, and then the tracked result of the tracker has been completed again? Is there any translation available?And from the train.py code provided in github, the model is only trained for a single dataset. And I found that in the data preprocessing, the preprocessing result itself is very large, with a storage space of about 6G. So for a task that needs to evaluate a single task containing multiple videos, such as hundreds of videos, and the video length is 0-20 seconds, is it necessary to perform very repetitive and tedious operations?I sincerely hope that you can answer as soon as possible. If you have any questions, please feel free to contact me！！