Method for performing anomaly detection

Abstract

The invention relates to a method (100) for performing anomaly detection, comprising the following steps: - Providing (101) sensor data, wherein the sensor data result from the acquisition of at least one sensor (2) of a vehicle (1), - Extracting (102) features of the sensor data, wherein the extracted features represent objects (3) in an environment of the vehicle (1), - Generating (103) bounding boxes based on the extracted features, wherein the bounding boxes represent regions in which the objects (3) are located in the environment of the vehicle (1), - Predicting (104) a respective class and a probability for the respective class for each of the bounding boxes, wherein the respective class indicates a type of object (3), - Determining (105) an anomaly value that characterizes a behavior and / or a position of a respective object of the bounding boxes.wherein the determination (105) is carried out on the basis of the extracted features, the generated bounding boxes and / or the predicted probability of the respective classes, - Performing (106) the anomaly detection on the basis of the determined anomaly value.