Visual odometry is the name used for algorithms that use vision to estimate the relative displacement of a mobile robot or sensor. Odometry accumulates the consecutive relative displacement to give a global estimation of the robot or sensor pose in the environment in respect to the initial pose, but bear in mind that it accumulates drift because the error on each relative displacement accumulates and grows without bounds. To solve this problem, a global localization or loop closure detection algorithm is needed. This is one of the components of a Visual SLAM system, but it also needs visual odometry to create a reference guess for the pose estimation.