STDyn-SLAM

STDyn-SLAM: A Stereo Vision and Semantic Segmentation Approach for SLAM in Dynamic Outdoor Environments

Authors: Daniela Esparza and Gerardo Flores.

License

STDyn-SLAM released under a GPLv3 license

Video

Install

Prerequisites

The necessary prerequisites are found in ORB-SLAM2 and DS-SLAM(https://github.com/ivipsourcecode/DS-SLAM):

• Pangolin

• OpenCV

• Eigen3

• DBoW2 and g2o

• ROS

• SegNet

• OctoMap and RVIZ

SegNet

Download and compile the SegNet package in /Examples/ROS/ORB_SLAM2_PointMap_SegNetM (https://github.com/TimoSaemann/caffe-segnet-cudnn5).

You also can download a new version of SegNet from (https://github.com/navganti/caffe-segnet-cudnn7). But you have to modify the root of SegNet in the CmakeLists.txt from /STDyn-SLAM, STDyn-SLAM/Examples/ROS/ORB_SLAM2_PointMap_SegNetM/ and STDyn-SLAM/Examples/ROS/ORB_SLAM2_PointMap_SegNetM/libsegmentation

OctoMap

You have to create a catkin workspace and download the OctoMap in /src

Building STDyn-SLAM

Change the name of the catkin workspace to yours.

cd catkin_ws/src
git clone https://github.com/DanielaEsparza/STDyn-SLAM
cd STDyn-SLAM
chmod +x BUILD.sh
./BUILD.sh

Run

Before executing, please run the next in the terminal. Modify ROOT_PATH by your container folder of STDyn-SLAM.

export  ROS_PACKAGE_PATH=${ROS_PACKAGE_PATH}:ROOT_PATH/STDyn-SLAM/Examples/ROS/ORB_SLAM2_PointMap_SegNetM

Real-Time Stereo

Execute the ros package of the stereo camera, modify the STEREO_RealTime.launch

cd src/STDyn-SLAM
roslaunch STEREO_RealTime.launch

KITTI Dataset

Modify the KITTI Dataset path in the KITTI.launch

cd src/STDyn-SLAM
roslaunch KITTI.launch

Own Sequence

Modify the path of your sequence in the PATH.launch

cd src/STDyn-SLAM
roslaunch PATH.launch