DynaSLAM Docker

This application builds DynaSLAM algorithm in a docker container.
It is required to have nvidia-docker installed.
DynaSLAM paper: https://arxiv.org/pdf/1806.05620.pdf
DynaSLAM git repository: https://github.com/BertaBescos/DynaSLAM

Build the docker

Run the following command to build the docker. Building the docker can take a while, and at least 15GB from disk space.

mkdir Dataset
sudo docker build -t dyna .

Run the docker

The following command will open a docker terminal

./Run_docker.sh

Access data from/to Docker

To move files from/to the docker just move to the Dataset folder (for example, the SLAM sequence dataset)

Run DynaSLAM

To run DynaSLAM a important note is to execute the terminal command at /root/DynaSLAM. Executions outside this folder will probably return Segmentation fault message.

./Examples/RGB-D/rgbd_tum Vocabulary/ORBvoc.txt Examples/RGB-D/TUM3.yaml /home/Dataset/rgbd_dataset_freiburg3_walking_xyz/ Examples/RGB-D/associations/fr3_walking_xyz.txt src/python/mask_rcnn_coco.h5

Exit the docker

Inside the docker just type the following command, but the docker will still be activate when you leave:

exit

If you want to return to the docker process:

./Attach_docker.sh

Kill docker process

Outside the docker, type the following command to show all docker process running

sudo docker ps -a

To kill a docker:

sudo docker rm <CONTAINER ID> 

Remove Docker image

Outside the docker, type the following command to show all docker images running

sudo docker images

To erase a docker image:

sudo docker rmi <IMAGE ID>