Sleep events observed in human sleep electroencephalogram (EEG) have a remarkable research value, and vertex sharp wave (VSW) is a common type of sleep event. However, there are currently no publicly available VSW annotation dataset or automatic detection method, limiting further research into the physiological mechanisms of VSW during sleep. Automatic detection algorithms require large dataset for training and validation. There is an urgent need of a VSW dataset and automatic VSW detection method. In this paper, we used the active learning (AL) method to construct a high-quality VSW annotation dataset of 1,167 VSW instances from 200 subjects involved in the Montreal Sleep Study Archive (MASS). The quality of the dataset was validated through four algorithms: fully convolutional network (FCN), multi-layer perceptron (MLP), support vector machine (SVM), and Transformer. Then we provided a baseline for VSW detection using FCN model. Based on this dataset, we further studied age difference in some physiological characteristics of VSW, including peak-to-peak value (PTP), root mean square (RMS), and spectrum power. This study establishes a publicly available VSW dataset and a VSW detection model, which provide a foundation for analyzing the physiological mechanisms of the VSW sleep event.