With the increase in ship traffic comes an increase in the likelihood of at-sea offences. Problems such as environmentally disastrous ship accidents, piracy, illicit fishing, drug trafficking, and illegal cargo movement can all be addressed by detecting ships as rapidly as feasible in satellite images. Automatic ship detection in remote sensing images is a challenging problem due to the complexity of scene clutter and the diversity of ship scale and position. In this study, we set up a pipeline with two models: a classifier for recognizing the presence of ships in images and a mask predictor for ship location. Because most images do not contain ships, they must pass through a binary classifier that predicts ships' existence. Subsequently, images containing ships undergo processing by the mask predictor, yielding a mask specific to each image. The proposed binary classification algorithm achieved benchmark results on the Airbus ship detection dataset with 98.26% accuracy, outperforming the scores obtained using traditional methods. The Cascaded Mask R-CNN network performance outperformed the Mask R-CNN, QueryInst, and DetectoRS networks based on mean average precision.