The goal of this work is to build an algorithm to automatically detect and segment ships in satellite images. A number of factors can make ship detection and segmentation a challenging task. Some of these factors include flaws in image quality like uneven brightness, obstruction, and also the fact that there are many images which have similar shape, color and texture. Another problem is that objects like islands, ports, whales etc look quite similar to ships. The algorithm had to be extremely accurate because lives and billions of dollars in energy infrastructure is at stake. In this paper we have used a custom MASK R CNN with backbone as ResNet 50 trained using MS COCO weights. Using mask shape of size 14*14, the algorithm is able to detect and segment ships with a Mean Average Precision of 0.605. The model takes approximately 30 seconds to segment 30 images. Hence this model can be used for real time maritime applications. © 2019, IJSTR.