- 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.
- Install Docker CE, if not already installed.
- Install Python 3.6 or later, if not already installed:
- Clone the Visual Road repository and build the benchmark image:
git clone https://github.com/georgia-tech-db/visualroad
cd visualroad
docker build -t carlasim/eva .- Generate a scale-one dataset with videos of 10 sec each with the following command. Note:
-lflag specifies what objects you need ground truth info for.
mkdir dataset
chmod 777 dataset
docker-compose run generator -s 1 -d 10 -l [car,pedestrian] dataset- The generator service supports a number of additional options (e.g.,
--height,--width,--duration). Executedocker-compose run generator -hfor a complete list.