cd
into project root directory and run
sudo docker-compose build
This takes a while.
cd
into project root directory and run
xhost +
followed by
sudo docker-compose up
Open a new terminal in the project root and type
sudo docker exec -it person_reid bash
This opens a bash prompt within the container.
To use a new set of input videos copy the video files into the "videos" folder
and update the option video_config["cams"]
in the file sfmt/config
accordingly.
To extract objects from a pair of videos in the "videos" folder run the command
python3 detection.py
in the previously opened interactive bash session within the container. For each time step of the videos a pickled file containing the detected objects will be created in the "detections" folder. The naming of the files follows the schema "step_x" indicating the time step of the contained detections.
The reid.py
contains a demo of how to use the extracted detections. Run it via
python3 reid.py
in the interactive bash session in the container.