A Hyperparameter Quality Assessment Method for UAV Object Detection Based on IER Rule

Unmanned aerial vehicles (UAVs) have been widely used in various target tracking applications through deep learning. The choice of the super parameters on the efficiency of model training and accuracy has important influence, however, the complexity of model structure and the difference of application background, the trainer cannot give precise hyperparameters. Therefore, A hyperparameter quality assessment method is proposed based on interval evidential reasoning (IER) rule in this paper. The interval confidence distribution is used to represent the interval probability generated by the uncertainty judgment, which improves the reliability of the evaluation. The evaluation process based on IER rule is defined, and the effectiveness of the evaluation is verified by experiments.