You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Would it be conceivably possible to impose constraints on tracklet construction in the tracking stage - for example, my objects getting bigger doesn't make any physical sense, but I see quite a bit of confusion in the tracking where objects get linked to larger objects in future timesteps so I'd like to be able to impose rules to make certain linkages (based on the visual features passed to the tracker) impossible or at least have very strong priors against. I was also wanting to be able to have variable search radius based on object attributes - in my case, larger objects are easier to discriminate and could therefore be confidently linked over much larger areas than small objects (where 'losing' a small object because it has travelled too far and re-initiating the track is also considerably less bad than 'losing' a large object that has travelled outside the search radius). I can imagine that these two could possibly be implemented through a similar pathway. Where might I go looking in the code to make a start?
The text was updated successfully, but these errors were encountered:
Would it be conceivably possible to impose constraints on tracklet construction in the tracking stage - for example, my objects getting bigger doesn't make any physical sense, but I see quite a bit of confusion in the tracking where objects get linked to larger objects in future timesteps so I'd like to be able to impose rules to make certain linkages (based on the visual features passed to the tracker) impossible or at least have very strong priors against. I was also wanting to be able to have variable search radius based on object attributes - in my case, larger objects are easier to discriminate and could therefore be confidently linked over much larger areas than small objects (where 'losing' a small object because it has travelled too far and re-initiating the track is also considerably less bad than 'losing' a large object that has travelled outside the search radius). I can imagine that these two could possibly be implemented through a similar pathway. Where might I go looking in the code to make a start?
The text was updated successfully, but these errors were encountered: