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
At the moment, detected objects are re-identified by their position. However, this approach has some drawbacks. For example, if two objects cross their way, their IDs might swap. It might be more reliable to determine the dominant color of an detected object and use it as a criteria to re-identify it (similar to color tracking #3). This might even be used to identify the object to track in the first place, e.g. only track a person with a red jacket. This might also be useful to filter false candidates (e.g. a tree that looks similar to a person in some frames).
The text was updated successfully, but these errors were encountered:
At the moment, detected objects are re-identified by their position. However, this approach has some drawbacks. For example, if two objects cross their way, their IDs might swap. It might be more reliable to determine the dominant color of an detected object and use it as a criteria to re-identify it (similar to color tracking #3). This might even be used to identify the object to track in the first place, e.g. only track a person with a red jacket. This might also be useful to filter false candidates (e.g. a tree that looks similar to a person in some frames).
The text was updated successfully, but these errors were encountered: