Overview: Working with only a limited amount of one dimensional, vector-based doppler radar data our client needed to increase the mean accuracy of pre-defined object classification. As the radar operates in airport vicinity, the classifier was to detect between a number of distinct airborne objects in the given dataset.

Technologies: Python, C++, XGBoost, libsvm, TensorFlow, signal processing, 1D convolutional autoencoders, SVM, SHAP

Objectives: Implementing a classifier working with one dimensional radar signal data. Create an easy to use and interpret pipeline for training and evaluation of results. Implement the proposed solution in C++ and run the classifier on a real time target hardware.