Home
The UCF SST CitySim dataset contains vehicle trajectories captured using drone videos. The videos were captured at 30 frames per second. Each row in the trajectory dataset represents a vehicle waypoint at a single frame in the video. Additionally, the dataset describes the rotation-aware vehicle bounding box in each row.
The following table describes the trajectory CSV columns and states their respective units.
| Name | Description | Unit |
|---|---|---|
| frameNum | Frame number of the vehicle waypoint captured at 30 frames per second | - |
| carId | Vehicle unique identifier that remains consistent for all vehicle waypoints across the entire video | - |
| carCenterX | Pixel x-coordinate of vehicle bounding box center point | Pixel |
| carCenterY | Pixel y-coordinate of vehicle bounding box center point | Pixel |
| headX | Pixel x-coordinate of vehicle bounding box front center point | Pixel |
| headY | Pixel y-coordinate of vehicle bounding box front center point | Pixel |
| tailX | Pixel x-coordinate of vehicle bounding box rear center point | Pixel |
| tailY | Pixel y-coordinate of vehicle bounding box rear center point | Pixel |
| boundingBox1X | Pixel x-coordinate of vehicle bounding box vertex 1 | Pixel |
| boundingBox1Y | Pixel y-coordinate of vehicle bounding box vertex 1 | Pixel |
| boundingBox2X | Pixel x-coordinate of vehicle bounding box vertex 2 | Pixel |
| boundingBox2Y | Pixel y-coordinate of vehicle bounding box vertex 2 | Pixel |
| boundingBox3X | Pixel x-coordinate of vehicle bounding box vertex 3 | Pixel |
| boundingBox3Y | Pixel y-coordinate of vehicle bounding box vertex 3 | Pixel |
| boundingBox4X | Pixel x-coordinate of vehicle bounding box vertex 4 | Pixel |
| boundingBox4Y | Pixel y-coordinate of vehicle bounding box vertex 4 | Pixel |
| carCenterXft | X-coordinate of vehicle bounding box center point in feet | Feet |
| carCenterYft | Y-coordinate of vehicle bounding box center point in feet | Feet |
| headXft | X-coordinate of vehicle bounding box front center point in feet | Feet |
| headYft | Y-coordinate of vehicle bounding box front center point in feet | Feet |
| tailXft | X-coordinate of vehicle bounding box rear center point in feet | Feet |
| tailYft | Y-coordinate of vehicle bounding box rear center point in feet | Feet |
| boundingBox1Xft | X -coordinate of vehicle bounding box vertex 1 in feet | Feet |
| boundingBox1Yft | Y -coordinate of vehicle bounding box vertex 1 in feet | Feet |
| boundingBox2Xft | X -coordinate of vehicle bounding box vertex 2 in feet | Feet |
| boundingBox2Yft | Y -coordinate of vehicle bounding box vertex 2 in feet | Feet |
| boundingBox3Xft | X -coordinate of vehicle bounding box vertex 3 in feet | Feet |
| boundingBox3Yft | Y -coordinate of vehicle bounding box vertex 3 in feet | Feet |
| boundingBox4Xft | X-coordinate of vehicle bounding box vertex 4 in feet | Feet |
| boundingBox4Yft | Y -coordinate of vehicle bounding box vertex 4 in feet | Feet |
| carCenterLat* | Global latitude of vehicle bounding box center point | Degrees |
| carCenterLon* | Global longitude of vehicle bounding box center point | Degrees |
| headLat* | Global latitude of vehicle bounding box front center point | Degrees |
| headLon* | Global longitude of vehicle bounding box front center point | Degrees |
| tailLat* | Global latitude of vehicle bounding box rear center point | Degrees |
| tailLon* | Global longitude of vehicle bounding box rear center point | Degrees |
| boundingBox1Lat* | Global latitude of vehicle bounding box vertex 1 | Degrees |
| boundingBox1Lon* | Global longitude of vehicle bounding box vertex 1 | Degrees |
| boundingBox2Lat* | Global latitude of vehicle bounding box vertex 1 | Degrees |
| boundingBox2Lon* | Global longitude of vehicle bounding box vertex 1 | Degrees |
| boundingBox3Lat* | Global latitude of vehicle bounding box vertex 1 | Degrees |
| boundingBox3Lon* | Global longitude of vehicle bounding box vertex 1 | Degrees |
| boundingBox4Lat* | Global latitude of vehicle bounding box vertex 1 | Degrees |
| boundingBox4Lon* | Global longitude of vehicle bounding box vertex 1 | Degrees |
| speed | Vehicle waypoint speed | Miles per Hour |
| heading | Vehicle waypoint heading relative to the global North | Degrees |
| course | Vehicle waypoint heading relative to the image coordinate X-axis | Degrees |
| laneId | Waypoint lane number according to the supplementary lane map | - |
*Only provided for US-based locations
The pixel coordinate system is displayed in the image below.
The vehicle center, head, tail, and bounding box vertices locations are described in figure below.
- Bounding box point 1 describes the front right vehicle corner
- Bounding box point 2 describes the rear right vehicle corner
- Bounding box point 3 describes the rear left vehicle corner
- Bounding box point 4 describes the front left vehicle corner
A lane map file (lanes.png) is provided for each location to describe the lane numbers.
Apart from the trajectory data, the CitySim dataset also provides simulation model for the corresponding location. The simulation model contains road network for microscopic traffic simulation software SUMO and automated driving software CARLA, which are both open-source simulators. The road network is built based on GIS data, and 3D high-fidelity architectures are embedded for the CARLA base map. Driving behavior model can be calibrated using the trajectories to produce realistic traffic flow in SUMO, and the calibrated model can be used for microscopic safety analysis or signal optimization. The CARLA base map enables researchers to conduct automated driving research including perception, CAV control in mixed traffic, etc.
Sumo is an open-source microscopic traffic simulation software that provides powerful functions for simulating traffic flow. The Sumo model released in the dataset contains the road network file (.net.xml file) of the video location and its surrounding vicinity.
CARLA is an open-source automated driving simulator built on Unreal Engine 4. The maps can be imported to CARLA to generate a digital world of the video location. It enables the researchers to conduct automated driving study including sensor simulation and CAV control algorithm, etc.
The signal data CSV file provides event level data. Each record in the signal data represents a change in one of the signals. The data is represented by eight different digits that corresponds to different signal phases at an intersection. The eight digits show whether the signal phase is red (r), yellow (y), or green (g). Furthermore, the event start time and duration were calculated based on the start frame and end frame corresponding to the recorded videos and trajectory data. The following table describes the signal CSV file columns and their respective units.
| Name | Description | Unit |
|---|---|---|
| startFrame | Signal event start frame | - |
| eventDuration | Signal event duration in seconds | Seconds |
| SBL | Southbound left turn signal phase (‘r’: red, ‘y’: yellow, ‘g’: green) | - |
| SBT | Southbound through signal phase (‘r’: red, ‘y’: yellow, ‘g’: green) | - |
| WBL | Westbound left turn signal phase (‘r’: red, ‘y’: yellow, ‘g’: green) | - |
| WBT | Westbound through signal phase (‘r’: red, ‘y’: yellow, ‘g’: green) | - |
| NBL | Northbound left turn signal phase (‘r’: red, ‘y’: yellow, ‘g’: green) | - |
| NBT | Northbound through signal phase (‘r’: red, ‘y’: yellow, ‘g’: green) | - |
| EBL | Eastbound left turn signal phase (‘r’: red, ‘y’: yellow, ‘g’: green) | - |
| EBT | Eastbound through signal phase (‘r’: red, ‘y’: yellow, ‘g’: green) | - |
A signal image file (signals.png) file is provided for each signalized intersection location to label the signals in the intersection.
The metadata CSV file contains information about the drone video recordings used to extract the vehicle trajectories. The following table describes the metadata CSV file columns and their respective units.
| Name | Description | Unit |
|---|---|---|
| fileName | File name | - |
| frameSizeX | Pixel x-coordinate frame size | Pixel |
| frameSizeY | Pixel y-coordinate frame size | Pixel |
| recordingFrameRate | Frame rate of the drone video used to extract the trajectories | Frames Per Second |
| totalFrames | Total number of frames in the file | - |
| recordingTime | Recording time of the file (hh:mm) | - |
| duration | Duration of the file | Seconds |
Each location contains 2 supplementary images: background.png and lanes.png. Signalized intersection locations additionally contain a signals.png file.
The background.png image is an aerial shot of the location with all the vehicles removed. It can be used for visualization purposes.
The lanes.png file describe the laneId values used in the trajectory CSV files.
The signals.png file describe the signals in the intersection.