how to keep the 3d boxes' size（h，w，z）in stabilization for one vehicle at different cameras detection, as well as in all ground?

[icefrogplus]

for one vehicle, there is only unique size (h, w, z) . So, how to keep same size at two questions:

1. at different camera's view.
2. at different moment since the vehicle moves on the ground.

[xiaosu-zhu]

Our tracking pipeline in the UAV view has introduced a few post-processings to make the annotation stablized. Under the assumption that the vehicles' size don't change over time, we assign the vehicle size of the same tracking ID with same size (by using the average value of all bbox of its trajectory).

As for different cameras, the size is inherint and transformed from the global 3D size, so we only need to keep the global size.

[icefrogplus]

Our tracking pipeline in the UAV view has introduced a few post-processings to make the annotation stablized. Under the assumption that the vehicles' size don't change over time, we assign the vehicle size of the same tracking ID with same size (by using the average value of all bbox of its trajectory).我们在 UAV 视图中的跟踪流程引入了一些后处理，使注释更稳定。假设车辆尺寸不随时间变化，我们为同一跟踪 ID 赋予相同尺寸的车辆尺寸（使用其轨迹中所有 bbox 的平均值）。

As for different cameras, the size is inherint and transformed from the global 3D size, so we only need to keep the global size.至于不同相机，尺寸是固有的，并从全局三维尺寸中转换而成，所以我们只需保持全局大小。

it makes sense, thanks!