AntiYH

> Correct! Thank you for your response. this is truly excellent work! In addition, I have another question. To date, have any deep learning-based multi-target tracking approaches demonstrated the ability...

> "Consistently maintain continuous target trajectories" is a somewhat arbitrary goalpost, but I would answer yes, at least in some contexts. Here's the first example that comes to mind: https://arxiv.org/abs/2105.03247...

> Correct! Dear Author, I have run the code you provided and noticed that although each target trajectory consists of 20 time steps, the prediction only focuses on the target...

> Regarding the first question: the whole trajectory is indeed utilized for predicting the final state, but we don't predict any intermediary states, correct. > > Regarding the second question:...

> Btw, regarding predicting intermediate states, we did some work in this direction: https://arxiv.org/abs/2312.17261 Yes, thank you very much for your reply. I have recently read the paper you recommended...