Smart Coastline Environment Management Using Deep Detection of Manmade Pollution and Hazards

A significant portion of the tourism industry and the fishing communities depend on the health of the seashores for their livelihood. Visual interpretation of this environment provides the required information for autonomous agents for decision support in large-scale operations such as smart beach management. Deep Neural Networks (DNN) are recently used for image classification and scene understanding with very good results. In this paper, a DNN is used for object detection in the scenes from seashore areas. This interpretation is used for decision support for various smart beach applications. The technique used for smart beach scene interpretation is the transfer learning on a pre-trained DNN using VGG architecture. The Single Shot Detector (SSD) technique is used for object detection in the collected dataset from the beach areas. A dataset from the beaches of the Caspian Sea is collected in order to provide an extensive simulation in the real-world setting. The experimental results showed that the accuracy of the presented technique is acceptable for various applications in the seashore environment.