'NN-ETT': CNN-based Extended Target Tracking

Extended Target Tracking utilizing Convolutional Neural Networks for extent estimation.

This repository contains code implementing the FUSION22 Paper

"CNN-based Shape Estimation for Extended Object Tracking using Point Cloud Measurements"
Simon Steuernagel, Kolja Thormann, and Marcus Baum

The three submodules can be freely exchanged. For example, a more elaborate (Extended) Kalman Filter could be implemented (e.g. in order to take into account a specific motion model), without impacting the implementation of the Shape Estimator.

The general functionality of the NN-ETT tracking architecture remains that of a sequentially updated filter.

This Repository

In this repository, all relevant code for the implementation of the NN-ETT tracker for single elliptical extended objects is implemented.

This includes trained weights for the Shape Estimation CNN, to be found here. The corresponding network implementation can be found here.

All other code can be found directly in the /src folder. To re-create the experiments presented in the paper, the file evaluation_plot_generation.py can be executed.

The implementation of the full NN-ETT tracking architecture (using the previously described CNN) can be found in the file nn_ett_tracker.py. Two implementations of reference algorithms, namely IAE and MEM-EKF* are included as well.

Finally, all other files in /src contain the remaining code used for data generation, visualization and general utility functions.