This work presents an application of combining physics-based and data-driven approaches for uncertainty quantification in the context of radar data analysis. We propose a methodology for estimating the quality of automotive radar using deep learning, which was initially trained on synthetic data from a complete radar signal processing chain simulation. The deep learning network learns the difference between a mid-resolution radar and a high-resolution radar, making it suitable for assessing detection quality. In the proposed methodology, the deep learning network operates in parallel with the physics-based radar detection algorithm, facilitating efficient uncertainty quantification of the detection quality. The additional information generated can be further used to enhance, for example, a Kalman filtering by refining the covariance matrices; thereby improving the results of the new process. By integrating both the physics-based algorithm and the data-driven model, our approach improves the detection capabilities of the system. This research contributes to the fusion of physics-based and data-driven approaches for uncertainty quantification in radar data analysis, with applications enhancing the results of algorithms that use the generated additional information.