ORB_SLAM2 with FCIS Segmentation

• Note: This repository is mainly built upon ORB_SLAM2 and FCIS. Many thanks for their great work.

Installation

Refer to the corresponding original repositories (ORB_SLAM2 and FCIS for installation tutorial).

Run

• First, run

python FCIS/experiments/slxrobot/segmentation_for_slam.py

• Then, run

./ORB_SLAM2/Examples/ROS/ORB_SLAM2/RGBD_Kinect ORB_SLAM2/Vocabulary/ORBvoc.bin ORB_SLAM2/Examples/RGB-D/kinect2_qhd.yaml

The current repository only supports RGBD sensor input.

After running, you can toggle Show DenseMap button to see the point cloud generated at the moment you pressed this button. And Show SegObjects will show the objects' type and their corresponding positions in the keyframes in which they exist.

Note

Please refer to the automap branch for ORB-SLAM2 with capability of automatically exploring the environment and build the map. And this branch also contains the code on how to save and reload map using boost serialization.