Only Numpy, Matplotlib and pytransform3d are required to be installed.

To visualize a viewing graph, please go to the directory rotation and play around with demo_test.ipynb. Some mapping methods between different rotation representations are implemented as well.

To check the effectiveness of the interior point method, please go to the directory rotation and find l1decoder.ipynb. You can do different tests by changing the size of matrix A and adding different level of perturbations. Note: the level of perturbations should not be too high.

To show the demo of camera location recovery (i.e. translation averaging), please go to the directory translation and directly run location_estimation.py so that a plot of recovered 3D locations will be immediately drawn. The plot will be exactly the Figure 2 from the Appendix of the report.

To re-produce the plot of MAE versus noise variance (i.e. Figure 3) in the report, modify robustness_test = 1 to robustness_test = 1 in the script location_estimation.py. Running this test would take some time.

The hyperparameters of the algorithms and the noise variance can be modified in the script location_estimation.py. To create new synthetic data, please run data_creation.py and feel free to modify the variables like number of cameras.