Semi-direct sparse odometry

video of running in rpg_urban dataset, and live video.

OpenCV 3.1.0 is used in this code.

sudo apt-get install libeigen3-dev

This code use the template implement of Sophus. Recommend the latest released version v1.0.0

This code use glog for logging.

sudo apt-get install libgoogle-glog-dev

Use Ceres-Slover to slove bundle adjustment. Please follow the installation page to install.

This code use Pangolin to display the map reconstructed. When install, just follow the README.md file.

After build and install, copy the file FindDBoW3.cmake to the directory cmake_modules

the Euroc Dataset is used for evalution. In the project directory, run the demo by the following commond

./bin/monoVO_euroc ./config/euroc.yaml ./calib/euroc_cam0.yaml dataset_path

note that the dataset_path is the directory contaion body.yaml

finally, there will be a trajectory.txt saved, and you can use the evo to evaluate.