Dear,
I came on your repo for your global data association based tracking approach. Since I'm thinking on working out a similar kind of tracking approach, I tried to setup your code and use it as a starting point. I did not succeed however. Could you provide a simple example, maybe based on publicly available material?
I tried myself using the TownCentre dataset for experimentation. I used the provided annotations as "detections" (all having the same score). To limit the size of the generated network, I experimented with only the first 50 images.
After some calculations I get the following result:
Optimal number of flow: 1
Optimal cost: 59912
Optimal flow:
{'source': {('00000051.jpg', 5, 'u'): 1}, ('00000051.jpg', 5, 'u'): {('00000051.jpg', 5, 'v'): 1}, ('00000051.jpg', 5, 'v'): {'sink': 1}}
which does not seem like MOT results 🤔
Dear,
I came on your repo for your global data association based tracking approach. Since I'm thinking on working out a similar kind of tracking approach, I tried to setup your code and use it as a starting point. I did not succeed however. Could you provide a simple example, maybe based on publicly available material?
I tried myself using the TownCentre dataset for experimentation. I used the provided annotations as "detections" (all having the same score). To limit the size of the generated network, I experimented with only the first 50 images.
After some calculations I get the following result:
which does not seem like MOT results 🤔