gehenhe / tracta Goto Github PK

Can you provide detailed installation steps or running environmental files？

Hi, thanks for the implementation code.

I'm trying to reproduce the result of the NLPRMCT dataset, and I met some problems.

1. How to get the similarity matrix? (same as #22)

• It seems that there is only SCT module and the rnmf algorithm module, but no the similarity measurement module.

2. How to set the parameters of the rnmf algorithm?

• I found some parameters in the rnmf algorithm that can be adjusted, I'm wondering how to change the parameters for the NLPRMCT dataset.

3. SCT algorithm with the non-overlapping camera scenario.

• It seems that the released code is implemented for the overlapping camera scenario, is the SCT algorithm for the non-overlapping camera scene a standard DEEPSORT algorithm?

Thank you!

Thanks for your great work! May I ask for the result files (e.g., json, txt, or others that are stored in an online driver) on each experimental dataset? I want to compare more details when I cite your work.

in config file __C.TEST.tracker_list, what value need to give ?
I considered two views and given __C.TEST.tracker_list = ['1', '2']

Am I wrong? Please suggest

Hello,

I enjoyed reading your paper. Do you plan to release the code for TRACTA in this repository as well?

Hi, thank you for sharing your hard work. I am trying to run your code but I have a problem. I get this error:

Traceback (most recent call last):
  File "tracking_multi_view.py", line 28, in <module>
    from detectron.pedestrain_det_msk import Pedestrain_Det_Msk
  File "/home/user/TRACTA/detectron/pedestrain_det_msk.py", line 26, in <module>
    from core.config import assert_and_infer_cfg
ModuleNotFoundError: No module named 'core'

I think it is part of the detectron package but there already exists a detectron package in the form of a folder in the project. Currently running code on:

Ubuntu 18.04.6
python 3.6.9

@GehenHe
Thanks for your wonderful work.
Can you provide a homography matrix for the EPFL dataset (Terrace, Passageway, Basketball) that is used for evaluating the model in the paper?
Thank you in advance

Hi. Thank you for publishing your implementation. I would like to get an explanation of certain parts of your implementation.

I have read through your paper and the implementation and I could not understand that the implementation of calculating similarity matrix S is consistent with the description in the paper.

Specifically, in your paper, you explain that you used an algorithm from another paper for function φ_{***} but I could not find one in that paper that would correspond to the implementation in this repository. I was also unable to locate where the inner division of the value of φ is performed.

Could you please provide an explanation regarding these?

Excuse me, referring to code, may I ask how to obtain the tracklets distance in the following cases:
(1) Two tracklets have no overlapping in the temporal space.
(2) Two tracklets are not aligned in the temporal space.

In case (1), when the distance may not be able to be calculated, how to make sure the K can be properly estimated and the assignment is correct?

Hi Author, I want to add yolov5 and OSNet for detection and person re-identification.

你好，首先感谢这项优秀的工作。
tracking_multi_view.py文件引用了top_view_visual，但是项目中没有找到这个包或者文件，也无法用pip安装，请问这个问题如何解决？

Hi, thank you for sharing your work. Is there an easy way to install caffe besides compiling from source?

Respected author for the result comparison with state-of-the-art method, can you please give details on dataset split you have followed for this work?

it will be a great help

Excuse me, may I ask if the application scenes are limited to pedestrians only? It seems that the body height is used to calculate the 3D positions. Given the bboxes, how to deal with sitting or lying persons?
code
code

如题，我想要在自己的视频上进行测试。但却无法正常运行。请问您能给一个详细点的instruction吗。