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
{{ message }}
Repository navigation
The track box get big width at image bottom when the moment target appear #456
As the target moves upward from the image bottom, the tracking box width gradually expands until the target is fully visible, then returns to its normal size. As the target moves downward from the lower boundary, the tracking box width gradually decreases from the moment the target enters the lower boundary. Could you tell me how to solve it?
You can refer: Vertical-Beach/ByteTrack-cpp#23
ByteTrack KF status is [cx, cy, a, h, vx, vy, va, vh], and width=a*h.
When the front of the car appears from the bottom of the screen, the initial shape is short and wide (that is, the aspect ratio a is very large).
If the initial value of the standard deviation of va is set very small, it is difficult to quickly change the aspect ratio.
The relevant parameter is in the stdArray within the Activate function of STrack.
Activity
BigDad1 commentedon Nov 12, 2025
I inferenced on tensorrt cpp in "/ByteTrack/tree/main/deploy/TensorRT/cpp/ ".
TRYOKETHEPEN commentedon Apr 7, 2026
You can refer: Vertical-Beach/ByteTrack-cpp#23
ByteTrack KF status is [cx, cy, a, h, vx, vy, va, vh], and width=a*h.
When the front of the car appears from the bottom of the screen, the initial shape is short and wide (that is, the aspect ratio a is very large).
If the initial value of the standard deviation of va is set very small, it is difficult to quickly change the aspect ratio.
The relevant parameter is in the stdArray within the Activate function of STrack.