Abstract
Using a public dataset of images of maritime vessels provided by Analytics Vidhya, manual annotations were madeon a subsample of images with Roboflow using the ground truth classifications provided by the dataset. YOLOv5,a prominent open source family of object detection models that comes with an out-of-the-box pre-training on theCommon Objects in Context (COCO) dataset, was used to train on annotations of sub-classifications of maritimevessels. YOLOv5 provides significant results in detecting a boat. The training, validation, and test set of imagestrained YOLOv5 in the cloud using Google Colab. Three of our five subclasses, namely, cruise ships, ROROs (RollOn Roll Off, typically car carriers), and military ships, have very distinct shapes and features and yielded positiveresults. Two of our subclasses, namely, the tanker and cargo ship, have similar characteristics when the cargoship is unloaded and not carrying any cargo containers. This yielded interesting misclassifications that could beimproved in future work. Our trained model resulted in the validation metric of mean Average Precision (mAP@0.5)of 0.932 across all subclassification of ships.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have