Need a tutorial for Vehicle tracking query (Q8)

[yangyueren]

Hello, I am attracted by the vehicle tracking query mentioned in the paper visualroad. I have several questions and hope to be answered.
I want to know all the vehicle tracks in the dataset, when does a vehicle pass by a camera, eg vehicle_id1 passes by camera_id3 during time 150s to 160s, vehicle_id1 passes by camera_id8 during time 230s to 225s. During the time 160s to 230s, vehicle_id1 is on the road so that no camera records it.

So I have the following questions:

1. During the Dataset Generator process, whether the tracks of each vehicle are recorded into a text file.
2. Whether the bounding boxes of each vehicle in every frame are recorded into a text file.

I would feel honored to get a reply.

[BrandonHaynes]

Hi Yang -- vehicle tracks are not currently recorded, but this would be a straightforward enhancement (by writing vehicle locations in the listener function within generator.py). I'd accept a PR for this and may have some time to add it as a feature in the near future.

Bounding boxes are not recorded, but the semantic videos indicate when cameras can see a vehicle. You can use these to formulate bounding boxes. However, this does not tell you which vehicle the camera is observing. If this is something that is needed, it would also require tracking additional metadata in the listener function.

Hope this helps! I'll leave this issue open for now as I'd like to incorporate one or both ideas into the generator.