A Dataset for Radar-based Traffic Gesture Recognition with Out-of-Distribution Detection

Radar-based gesture recognition has received a lot of attention in recent years, with many applications and neural network architectures proposed. However, public datasets have to be available for the reliable design and validation of algorithms, as is the case, for example, in computer vision. Hence, this paper presents a comprehensive traffic gesture dataset with different radar data representations for eight traffic gestures, recorded for 35 participants. Beyond the traffic gestures, the dataset comprises sequences of gesturing and unknown motions for the evaluation of continuous classification and Out-of-Distribution (OOD) detection, respectively. Results are presented for classification accuracy and OOD detection, serving as a baseline for further research. The dataset and additional resources available for download at https://www.uni-ulm.de/en/in/mwt/research/online-database/traffic-gesture-dataset