Visual Road

A Video Data Management Benchmark

Read the Paper
Get the Benchmark 

About Visual Road

Video database management systems (VDBMSs) have recently re-emerged as an active area of research and development. To accelerate innovation in this area, we present Visual Road, a benchmark that evaluates the performance of these systems. Visual Road comes with a dataset generator and a suite of benchmark queries over cameras positioned within a simulated metropolitan environment. Visual Road's video data is automatically generated with a high degree of realism, and annotated using a modern simulation and visualization engine. This allows for VDBMS performance evaluation while scaling up the size of the input data.

Visual Road is designed to evaluate a broad variety of VDBMSs: real-time systems, systems for longitudinal analytical queries, systems processing traditional videos, and systems designed for 360◦ videos. Visual Road relies on the Unreal Engine for physical simulation and rendering, and the Carla simulator as a back-end engine (including its assets, geographic elements, and actor automation logic).

Visual Road architecture diagram


Sample Traffic Camera Videos

The following videos are representative of the traffic cameras found in a synthetic Visual Road dataset:

Rain with light traffic
Postpluvial with heavy traffic
Postpluvial with heavy auto and foot traffic
Sunny with moderate foot traffic

Pregenerated Datasets

Name Scale Resolution Duration Version Configuration Download
1K-Short-1 1 1K (960x540) 15 min 1 Link 1k-short.tar.gz
1K-Short-2 2 1K (960x540) 15 min 1 Link 1k-short-2.tar.gz
1K-Long-4 4 1K (960x540) 60 min 1 Link 1k-long-4.tar.gz
2K-Short-2 2 2K (1920x1080) 15 min 1 Link 2k-short-2.tar.gz
4K-Short-1 1 4K (3840x2160) 15 min 1 Link 4k-short.tar.gz

Share your dataset configuration!

If you’ve generated a dataset and would like to share its configuration with the world, please post it here with the details! To view a list of dataset configurations, please click here.

Related Publications


This work is supported by the NSF through grants CCF-1703051, IIS-1546083, CCF-1518703, and CNS-1563788; DARPA award FA8750-16-2-0032; DOE award DE-SC0016260; a Google Faculty Research Award; an award from the University of Washington Reality Lab; gifts from the Intel Science and Technology Center for Big Data, Intel Corporation, Adobe, Amazon, Facebook, Huawei, and Google; and by CRISP, one of six centers in JUMP, a Semiconductor Research Corporation (SRC) program sponsored by DARPA.