DS_SLAM is based on DSO (https://github.com/JakobEngel/dso). I add the loop closing and make it as complete vSLAM system. Please have a look DSO's README (https://github.com/JakobEngel/dso/blob/master/README.md), and compile and run the system.
System Dependencies:
- DSO dependencies
- Ceres Solver (http://ceres-solver.org/installation.html)
Install dependencies and compile the system:
cd DS_SLAM
mkdir build
cd build
cmake ..
make -j3
I run example on three dataset:
- TUM-Mono: https://vision.in.tum.de/mono-dataset
- Kitti: http://www.cvlibs.net/datasets/kitti/eval_odometry.php
- EuRoC: https://projects.asl.ethz.ch/datasets/doku.php?id=kmavvisualinertialdatasets
run ds_slam_dataset, config DBoW file like "dbow=xxx" (file is under Examples/DBoW), other configuration is same as DSO.
Improved points include:
- Add loop closing.
- Povide loop detection with two feature, BRIEF and ORB.
- Support two modes of loop closing and vo, separation mode and coupling mode.
- Improve display fo pose graph.
- VINS-Mono (https://github.com/HKUST-Aerial-Robotics/VINS-Mono)
- LDSO (https://github.com/tum-vision/LDSO)
DS_SLAM, like reference SLAM system, is licensed under GPLv3. Dependent third-party libraries include MIT, DSB, GPL etc.