A Deep Convolutional Neural Network Applied to Ship Detection and Classification

The task of detecting a ship is relevant for a variety of applications in both military and civilian fields, from maritime traffic surveillance to sea pollution monitoring. Despite the recent significant attention, deep learning-based object detection algorithms have received, they are still rarely applied in the detection of ships. Expectedly, we see even fewer applications when the goal is to detect specific ship classes, although those could deliver extra valuable information. In this work, we build a ship detection algorithm capable of distinguishing six common ship types: ore carrier, bulk cargo carrier, general cargo ship, container ship, fishing boat and passenger ship. To achieve this objective, we firstly built a deep convolutional neural network in python, taking advantage of the TensorFlow framework. Secondly, we analyzed our model’s ability to generalize on a new set of images, apart from the training and testing sets. The results showed that our proposed model is close to state-of-the-art performance since it was able to perform well on the test set - mAP = 97.62%. There is still room for improvement on the model robustness, associated primarily with possible training set limitations. In practice, this paper will contribute to the advance of research and applications on ship detection.

PDF

Datasets


  Add Datasets introduced or used in this paper

Results from the Paper


  Submit results from this paper to get state-of-the-art GitHub badges and help the community compare results to other papers.

Methods