search

obendidi / Tracking-with-darkflow

Real-time people Multitracker using YOLO v2 and deep_sort with tensorflow
GNU General Public License v3.0
523 stars 174 forks source link

ID Numbers #3

Open arevel190 opened 7 years ago

arevel190 commented 7 years ago

Hi,

I would like to know if there is a way to mitigate the problem of getting the IDs to jump to much from number. I am currently using your code and whenever a new person enters the frame the ID number tends to jump at least 10 numbers up.

Thanks!

obendidi commented 7 years ago

I'll reopen this because I'me very interested to see if anyone has any idea or advise on how to fix this issue and increase accuracy of tracking in general, I'm thinking a good way to start is to test using multiple object detectors(RFCN, Faster RCNN, ....) and compare the tracking with deep sort from each of them to see how the detection can influence the tracking if at all

nidhinkk commented 6 years ago

I am also facing the issue with the ID number. The increment number is not providing all the time. If at a time 3 person are present the adjacent numbers will be provided. But if a new person emerges in the next frame the Id may be a larger value. Does anyone find a solution?

ParticularJ commented 6 years ago

@bendidi Hello, your method is awesome! But, I am also facing the issue with the ID number.Do you have good ways to solve it? Thanks.