UNESCO estimates that our planet’s oceans and lakes are home to more than three million shipwrecks. Of these three million, the locations of only 10% are currently known. Apart from the historical and archaeological interest in finding wrecks, there are other reasons why we need to know their precise locations. While a shipwreck can provide an excellent habitat for marine life, acting as an artificial reef, shipwrecks are also potential sources of pollution, leaking fuel and corroding heavy metals. When a vessel runs aground on an iron-free environment, changes in the chemistry of the surrounding environment can occur, creating a discoloration called black reef. In this work, we examine the use of supervised deep learning methods for the detection of shipwrecks on coral reefs through the presence of this discoloration using satellite images. One of the main challenges is the limited number of known locations of black reefs, and therefore, the limited training dataset. Our results show that even with relatively limited data, the simple eight-layer, fully convolutional network has been trained efficiently using minimal computational resources and has identified and classified all investigated black reefs and consequently the presence of shipwrecks. Furthermore, it has proven to be a useful tool for monitoring the extent of discoloration and consequently the ecological impact on the reef by using time series imagery.