This repo provides a Python implementation of a basic monocular visual odometry algorithm using various feature detectors. This program was part of a final project for one of my classes (see report here).
Credit to Avi Singh, whose C++ code was referred to extensively for this program.
Before running, be sure to visit the Kitti Odometry Benchmark data set and download the color odometry data set (I used color, but the grayscale set should work with minor changes to the code) and odometry ground truth poses. Then, in visualodom.py, near the beginning, set the value of path to the path to the folder containing the two downloaded datasets (I assumed both are in the same folder).
Demo running program on Sequence 09 in Kitti data set at 11x speed.