This repo is an extension of visualroad, please check their project website for more details about the benchmark, links to the papers, sample videos, and pregenerated datasets.
git clone https://github.com/georgia-tech-db/visualroad
cd visualroad
docker build -t carlasim/eva .
Synthetic Dataset Generation
Generate a scale-one dataset with videos of 10 sec each with the following command. Note: -l flag specifies what objects you need ground truth info for.
The generator service supports a number of additional options (e.g., --height, --width, --duration). Execute docker-compose run generator -h for a complete list.