FMCW radar systems are increasingly used in diverse applications, and emerging technologies like JCAS offer new opportunities. However, machine learning for radar faces challenges due to limited application-specific datasets, often requiring advanced simulations to supplement real-world data. This paper presents a setup for generating synthetic radar data for indoor environments, evaluated using CNNs. The setup involves comprehensive modeling, including far-field antenna simulations, variations in human radar cross-section, and detailed representations of indoor environments with their corresponding propagation channel properties. These synthetic data are used to train CNNs, and their performance is assessed on real measurement data. The results demonstrate that CNNs trained on synthetic data can perform well when tested on real measurement data. Specifically, the models trained with synthetic data showed performance comparable to models trained with real measurement data, which required a minimum of 300 samples to reach similar levels of accuracy. This result demonstrates that synthetic data can effectively train neural networks, providing an alternative to real measurement data, particularly when collecting sufficient real-world samples is difficult or costly. This approach significantly reduces the time required for generating datasets, and the ability to quickly label data in simulations simplifies and accelerates post-processing. Additionally, the generated datasets can be made more heterogeneous by introducing varying signal conditions, enhancing the diversity and robustness of the training data.

Low resolution, sparse reflections, and environmental noise limit the reliability of radar-based object detection. This letter presents a YOLO-inspired deep learning model with dual-radar fusion to enhance detection robustness. Range-Doppler maps from two static 60 GHz FMCW radars are processed using a dual-backbone architecture with CBAM-based attention and a lightweight dynamic weighting module. The system monitors moving humans in a parking garage. At the best operating point, the proposed fusion improves the F1-score from 0.944 (single radar) to 0.962, with precision/recall increasing from 0.930/0.959 to 0.953/0.972. At matched recall ($\approx 0.967$), the false positive rate decreases from 0.070 to 0.031, corresponding to a reduction of about 55%. Real-time performance is maintained with inference speeds above 100 FPS on a desktop CPU. These results demonstrate that dual-radar feature fusion enables accurate and efficient radar perception in cluttered environments.