nhmduy / lmgp Goto Github PK

非常感谢你们的工作，静待代码；

Thanks for your nice work! May I ask a question as to how to obtain detections in this paper?

Ignoring the tracking performance, I find your detection of Table 1 is much better than previous works. However, as I know, current sota multi-view detection still cannot generate so good detections on WILDTRACK.

In your paper, it was mentioned
"We have demonstrated that nearly perfect multiple object tracking results can be obtained in crowded scenes given the right conditions (as for WILDTRACK)"
-> do you mean that GT detections are used for the tracking?

Hi, thanks for your great work!

I wonder what train-test split ratio used in each dataset.
In Wildtrack, many papers including yours follow the same standard, which is 90% for training and 10% for testing.
How about the other dataset like Campus and PETS09?

Thanks for your kind reply in advance.

Thank you for your great work.

I am wondering if you can give me a hint about when you will release the code?

Thank you in advance.