Method for detecting at least one object in surroundings by means of reflection signals of a radar sensor system

A method for training a neural network used for detecting at least one object using reflection signals of a radar sensor system, comprising:

• providing respective spatial coordinates of a multiplicity of reflection signals of at least one frame of the radar sensor system;
• providing respective first measurement attributes of at least a first portion of the multiplicity of reflection signals of the at least one frame of the radar sensor system;
• transforming the respective spatial coordinates of the multiplicity of reflection signals of the at least one frame into an occupancy grid;
• providing spatial coordinates of the at least one object to be detected by the reflection signals of the radar sensor system;
• generating the occupancy grid, wherein the multiplicity of the reflection signals of the at least one frame are spatially represented in the occupancy grid by transforming the respective spatial coordinates of the multiplicity of the reflection signals of the at least one frame into the occupancy grid and assigning the respective first measurement attribute of the multiplicity of the reflection signals of the at least one frame to the transformed spatial coordinates of the multiplicity of the reflection signals of the at least one frame;
• generating an input tensor using the occupancy grid that spatially represents the multiplicity of the reflection signals of the at least one frame and the respective assigned first measurement;
• forming a multiplicity of input tensors generated in this manner, each having a multiplicity of different reflection signals of at least one frame of the radar sensor system;
• and
• training the neural network using the multiplicity of input tensors and respective assigned provided spatial coordinates of the at least one object to be detected, the neural network being trained to output both an object classification and a location of the object as output based on a respective input tensor received as input.