SVO 2.0 is a robust semi-direct SLAM by Prof. Davide Scaramuzza and the teams from UZH and ETH Zurich that works great for fisheye cameras. While it is not open-source, the teams provide binaries for Ubuntu that run on ROS. The complete algorithm includes bundle adjustment, but the binaries only include place recognition without bundle adjustment.
With this short walkthrough you will be able to download, install and run SVO 2.0.
You can also visit their page at: http://rpg.ifi.uzh.ch/svo2.html
Also for further reading, see the publication:
Christian Forster, Zichao Zhang, Michael Gassner, Manuel Werlberger, Davide Scaramuzza; SVO: Semi-Direct Visual Odometry for Monocular and Multi-Camera Systems; IEEE Transactions on Robotics, Vol. 33, Issue 2,April 2017
Video example of SVO 2.0: SVO 2.0: Semi-Direct Visual Odometry for Monocular and Multi-Camera Systems - YouTube