Internal waves (IWs) are a widely occurring phenomenon in the global oceans. They have significant impacts on marine environments through their propagation mechanisms. In this study, we collect 390 satellite images with 775 labels from Sentinel‐1 Synthetic Aperture Radar (SAR) from 2014 to 2021 to construct a data set containing IW packets in Andaman Sea, South China Sea, Sulu Sea, and Celebes Sea, respectively. All the SAR images acquired are pre‐processed to provide clear IW packets for better visualization. To better detect IWs in global oceans automatically, a machine‐learning based model is proposed to focus on different channels and the spatial information. The precision, recall, and mean average precision of our improved model applied on the IW data set is up to 98.7%, 96.9%, and 98.9%, respectively. Various cases of IWs images are analyzed to illustrate detection quality in different stripes, scales, and propagation directions of IWs. The experimental results indicate that our data set is helpful to better detect IWs, and proposed network can accurately detect IWs in SAR images under various conditions.