A method for initializing a tracking algorithm for target objects, includes generating a 3D point cloud of the target object and iteratively determining a spatial position and orientation of the target object using a 3D model. A spatial position and orientation of the target object is first determined using an artificial neural network, thereafter the tracking algorithm is initialized with a result of this determination. A method for training an artificial neural network for initializing a tracking algorithm for target objects includes generating a 3D point cloud of the target object by a scanning method, and iteratively determining a spatial position and orientation of the using a 3D model of the target object. The artificial neural network is trained using training data to initially determine a spatial position and orientation of the target object and thereafter initialize the tracking algorithm with a result of this initial determination.