Dataset Support

This page visualizes sample traffic scenarios of every dataset that Tactics2D supports. If you want to propose new datasets to be supported by Tactics2D, please raise an issue.

HighD

HighD is a dataset collected by Aachen University, German. This dataset "includes post-processed trajectories of 110 000 vehicles including cars and trucks extracted from drone video recordings at German highways around Cologne during 2017 and 2018. At six different locations, 60 recordings were made with an average length of 17 minutes (16.5 h in total) covering a road segment of about 420 m length. Each vehicle is visible for a median duration of 13.6 s."

An insight of HighD can be found in SCP-CN-001/trajectory_dataset_support. For those interested in accessing the complete dataset, please follow this link.

Location 1
Location 2
Location 3
Location 4
Location 5
Location 6

InD

InD is a dataset collected by Aachen University, German. This dataset "contains more than 11500 road users including vehicles, bicyclists and pedestrians at intersections in Germany. The dataset consists of 10 hours of measurement data from four intersections".

An insight of InD can be found in SCP-CN-001/trajectory_dataset_support. For those interested in accessing the complete dataset, please follow this link.

Location 1
Location 2
[TODO] Location 3: The annotation of location 3 will be improved later.
Location 4

RounD

RounD is a dataset collected by Aachen University, German. This dataset "contains a total of six hours of recordings with more than 13,746 road users including cars, vans, trucks, buses, pedestrians, bicycles and motorcycles" from three roundabouts.

An insight of RounD can be found in SCP-CN-001/trajectory_dataset_support. For those interested in accessing the complete dataset, please follow this link.

Location 0
Location 1
Location 2

ExiD

ExiD is a dataset collected by Aachen University, German. This dataset contains "92 separate recordings (more than 16 hours of video) with traffic at seven different measurement locations at entries and exits of the German Autobahn...The data also contains variations in form of different speed limits on the highways: four locations have no speed limit, two locations have a speed limit of 100 km/h and one location has a speed limit of 120 km/h."

An insight of ExiD can be found in SCP-CN-001/trajectory_dataset_support. For those interested in accessing the complete dataset, please follow this link.

Location 0
Location 1
Location 2
Location 3
Location 4
Location 5
Location 6

Argoverse 2

Argoverse 2, curated by Argo AI, is sourced from six prominent U.S. cities, namely Austin, Detroit, Miami, Pittsburgh, Palo Alto, and Washington, D.C. The motion forecasting dataset "encompasses 250,000 scenarios featuring trajectory data" spanning across 10 distinct object types. "Each scenario, lasting 11 seconds, incorporates the 2D, birds-eye-view centroid, and heading information of each tracked object, sampled at 10 Hz." In its entirety, the dataset spans 763 hours and covers a distance of 2110 km.

An insight of Argoverse 2 can be found in SCP-CN-001/trajectory_dataset_support. For those interested in accessing the complete dataset, please follow this link.

Dragon Lake Parking (DLP)

DLP dataset is collected by University of California, Berkeley, CA, USA. "The entire DLP dataset contains 30 scenes, 317, 873 frames, 5, 188 agents, and 15, 383, 737 instances".

An insight of DLP can be found in SCP-CN-001/trajectory_dataset_support. For those interested in accessing the complete dataset, please follow this link.

INTERACTION

INTERACTION dataset is collected by University of California, Berkeley. "In the roundabout scenarios, 10479 trajectories of vehicles from five different locations were recorded for around 365 minutes. Similarly, in the unsignalized intersection scenarios, three locations were included and 14867 trajectories were collected for around 433 minutes. In the merging and lane change scenarios, 10933 trajectories were recorded at two locations for around 133 minutes. Finally, one location was selected for the signalized intersection, which provided 3775 trajectories for around 60 minutes."

An insight of INTERACTION can be found in SCP-CN-001/trajectory_dataset_support. For those interested in accessing the complete dataset, please follow this link.

DR_BGR_Intersection_VA
DR_CHN_Merging_ZS
DR_CHN_Roundabout_LN
DR_DEU_Merging_MT
DR_DEU_Roundabout_OF
DR_USA_Intersection_EP0
DR_USA_Intersection_EP1
DR_USA_Intersection_GL
DR_USA_Intersection_MA
DR_USA_Roundabout_EP
DR_USA_Roundabout_FT

NuPlan

NuPlan is the world's first large-scale planning benchmark for autonomous driving. The data is recorded over 4 cities - Boston, Pittsburgh, Singapore and Las Vegas. The nuPlan v1.0 dataset consists of over 15,000 logs and 1300+ hours of driving data. The nuPlan v1.1 dataset brings multiple improvements over the v1.0 dataset. Meanwhile, some log files are removed to guarantee the accuracy and reliability of the dataset.

An insight of NuPlan can be found in SCP-CN-001/trajectory_dataset_support. For those interested in the complete dataset, please follow this link.

NuPlan also has its own visualizing toolkit package at nuplan-devkit. The documentation of the devkit is here.

[TODO]: Fix the error in parser.

Waymo Open Dataset v1.2 (WOMD)

Waymo Open Motion Dataset is collected by Waymo from six locations: San Francisco, Phoenix, Mountain View, Los Angeles, Detroit, and Seattle. This dataset has "103,354, 20s 10Hz segments (over 20 million frames)" in v1.2 version, equivalent to approximately 574 hours of rich driving scenarios. Each segment has its own 3D map data.

An insight of WOMD can be found in SCP-CN-001/trajectory_dataset_support. For those interested in the complete dataset, please follow this link.

[TODO]: In the current map description of WOMD, the lane components are removed. We plan to add a function to detect and construct lanes from the map automatically.