A pipeline for monocular visual odometry
This is the way to install within a virtual environment created by
using pipenv:
$ pipenv install
$ pipenv shell
$ cd lib-visualodo
$ pip install -r requirements.txt
$ python setup.py develop --no-deps
To perform the unit tests present run:
$ make -C lib-visualodo tests-clean tests
The output is generated in the folder in lib-visualodo/out-comptests/.
In case you want to enable the plotting features (i.e. image containing the used mask, the feature extraction and matching or the outlier filtering), set the relevant variable to True in lib-visualodo/src/duckietown-visualodo/data/default.yaml
Detailed documentation for running the demo and on the implementation of the visual odometry pipeline can be found on the Duckiebook
The package is structured the following way: in ros-visualodo, the code responsible for the node instantiation and the communication with other nodes is placed. In lib-visualodo, all the actual logic is implemented, the parameters are contained and the unit tests are present.