To train and evaluate our enhancement framework, SS2DS, we establish a benchmark for the spike stream enhancement. The benchmark includes a synthetic spike stream enhancement dataset, SED, and a real sparse spike streams dataset for driving scenarios, SSDD. The SED dataset consists of 125 randomly high-speed dynamic scenes where 100 scenes are as the training set, named SED$_{tr}$, and 25 scenes are as the test set with ground truth, named SED$_{te}$.  This dataset consists of 24 driving scenes and each scenario records a spike stream with 20,000 frames.