SOTA-SLAM

TODOs (in the order of priority)

SIFT features (a bit later with CUDA support)
camera model independent (pinhole, fisheye and perspective at first)
Inverse depth parametrization
Voxel-based Map
GTSAM as pose graph backend
IMU fusing
mix of direct and indirect approach
online camera calibration
multicameras (follow RPG in implementation)
ROS wrapper

Tasks

Simple Visual Odometry pipeline (with thoughts about future upgrade to vSLAM): feature extraction, matching, pose estimation through Essential Matrix estimation, 3D visualization (pangolin)

Notes on some implementation details

c++-17
clang + its stdlib
ninja instead make (google, at first)
NO RAW POINTERS