search

cheind / py-motmetrics

:bar_chart: Benchmark multiple object trackers (MOT) in Python
MIT License
1.36k stars 258 forks source link

would you please suggest code for multi camera multi object tracking evaluation #157

Open ssbilakeri opened 2 years ago

Hasankanso commented 2 years ago

check mcta metric in http://www.mct2014.com/Evaluation.html

it is implemented in matlab...I wish I can find it in python!

angelcarro commented 2 years ago

You can also use this library for this purpose by changing the frame index the following way: frame_ind = frame_ind*1000 + cam_number This way you can merge the results from all the cameras and use this library.

SajjadPSavoji commented 1 year ago

You can also use this library for this purpose by changing the frame index the following way: frame_ind = frame_ind*1000 + cam_number This way you can merge the results from all the cameras and use this library.

@angelcarro Can you be more detailed? I'm assuming that you are trying to merge results(from multiple camera) by assigning new frame numbers to each source in a way that videos from different sources do not overlap in time.

What does 1000 represents in your solution?

angelcarro commented 1 year ago

Yes, the idea is to generate a new frame index to merge the information from the multiple cameras without collisions. This operation is performed both to the ground truth annotations and to the tracking results.

1000 is just a number bigger than the number of cameras. Using 1000 (or 100) is useful to visually separate the frame index and the camera number.