Whether for scientific research or commercial use, eye tracking technology is an important tool in many areas of society. Despite its wide range of applications, this technology has not yet been widely popularized. In order to facilitate the research and promotion of eye tracking technology, the GazeCapture dataset was created. As the world's first large-scale dataset for eye tracking technology, GazeCapture contains 2.5 million frames of data from more than 1,450 volunteers. Through the application of this dataset, the convolutional neural network iTracker used for eye tracking research has been fully trained, and the error has been greatly reduced compared to previous methods.