A Vision-Based Application for Container Detection in Ports 4.0.

The recent crises, such as COVID-19 or the Ukrainian war, have reinforced the need for the logistics industry to permanently optimize its operational processes for remaining competitive at the global level. In that sense, maritime logistics play a vital role, making them essential for modern society. The overall objective of this work is framed within the optimisation of maritime terminal yard operations through the application of computer vision services under the auspicious of Industry 4.0. In particular, the work focuses on object detection of containers in a maritime port terminal environment. Our study shows that a trained ResNet50 convolutional neural network on container detection within a maritime port obtains a total loss (a validation metric for object detection models) lower than one, as recommended for reliability in the posterior detection phase. This result testifies to the potential use of this algorithm for container detection in the context of maritime terminals, helping on automating certain manual processes for port operators.