As mentioned in Section 3.3, precision and recall values can be influenced by the threshold for the overlap ratio a0, and are varying according to the threshold. Therefore, the precision-recall curve of the model is computed by setting different threshold values during detection. As the area under the precision-recall curve (AUC) reflects the detection accuracy, i.e. higher curve means higher accuracy, the curves are plotted to evaluate the models in experiment 2 and 3.