Awesome
D2 TPred : Discontinuous Dependency for Trajectory Prediction under Traffic Lights
How was the data collected?
The data in VTP-TL comes from at urban intersections with traffic lights is used to predict vehicles trajectory in different times of day and provides a broad range of real-world driving scenarios. We use drones to hover at 70 to 120 meters above the traffic intersections, as statically as possible, to record vehicle trajectories passing through the area with a bird’s-eye view in the daytime of the non-rush hours, rush hours, and the evening.
<div align=center> <img src="https://github.com/VTP-TL/D2-TPred/blob/main/drone.png" width="780" height="312" alt=" "/><br/> </div>Where was the data collected?
We choose 3 different traffic intersections, including crossroad, T-junction, and roundabout scenarios. In these scenario, they own the different number of roads and traffic lights, and cause to different movement behaviors for vehicles.
<div align=center> <img src="https://github.com/VTP-TL/D2-TPred/blob/main/scenarios.png" width="762" height="628" alt=" "/><br/> </div>Summary of the Dataset
In the VTP-TL dataset, we have collected data from 3 different categories of traffic scenarios using drones. The summary of the data is listed in the following table.
<div align=center> <img src="https://github.com/VTP-TL/D2-TPred/blob/main/summary.png" width="772" height="503" alt=" "/><br/> </div>Included Materials
For the 3 recording scenarios, we include 2 files for each scenarios:
- The sample of video clips (xxx.mp4)
- Recorded vehicle trajectory file (xxx.txt) where, we provide trajectories information in pixel.
Recorded Vehicle Trajectory files (xxx.txt)
F_id: column 1. For each agent (per Agent_id), frame_id represents the frames the agent appears in the video.
A_id: column 2. For each xxx.txt file, the Agent_id starts from 0, and represent the ID of the agent.
x: column 3, the x position of the agent at each frame. The unit is pixel.
y: column 4, the y position of the agent at each frame. The unit is pixel.
Lane_id: column 5, For each xxx.txt file, the Lane_id starts from 0, and represent the ID of the traffic lane.
pa: column 6, For each xxx.txt file, the inperception is set as 0 or 1, and represent whether vehicle locates in the influencing area of traffic light.
f: column 7, For each xxx.txt file, the isfirstobj is set as 0 or 1, and represent whether vehicle is the first agent in the influencing area of traffic light.
Lig_id: column 8, For each xxx.txt file, the Lig_id starts from 0, and represent the ID of the traffic light.
ls: column 9, For each xxx.txt file, the Ls is set as 0, 1 and 2, and represents the state of traffic light.
mb: column 10, For each xxx.txt file, the Mb is set as 0, 1 and 2, and represents the movement behaviors of vehicle.
lt: column 11, For each xxx.txt file, the Ldurtime represents the durtime of traffic light.
Example:
<div align=center> <img src="https://github.com/VTP-TL/D2-TPred/blob/main/smaple.png" alt=" "/><br/> </div>