Open NeerajKomuravalli opened 4 years ago
Mahalanobis distance: gating_distance() in kalman_filter.py
appearance: gated_metric() in tracker.py calls self.metric.distance(features, targets)
, which is the cosine distance between the appearance vectors
As mentioned in the research paper assignment problem is solved with the help of two metrics, Mahalanobis distance of track and the detected bounding box and cosine distance of the appearance vector but when I looked into the code implementation I couldn't find the part where Mahalanobis distance is computed but square distance is computed instead. Can I know why was that done?
PS: I am really sorry if it was implemented but I missed it, in that case can you point me to part of the code where it's been implemented.