Object detection under heavy rain conditions is a challenging task that has been addressed numerous times by methods that aim to remove the ill effects caused by rain particles. While these approaches are successful in progressing the state-of-the-art in de-raining, the notion of rainy data being directly useful in object detection is frequently over looked. In this paper we discuss the usefulness of rainy data in improving the robustness of object detectors in rainy conditions. In the process, we ascertained that detector models trained on rainy data significantly outperformed models that were trained on only clear images. We also identify that the rear lights of vehicles are powerful features that may be used to enhance the detection and tracking of vehicles. This was identified by the propagation of these features to upper layers of the network, when other distinguishing features, such as the outlines of the car were less successful in progressing up the model. Experimental results, although still slightly underperforming, were a proof of concept that the considerations in this paper do improve the overall accuracy of detection models.