Abstract

In this chapter we present efforts on convolutional and transformer-based deep learning neural networks for the automated detection of corrosion on pipes and inside of vessels in the Oil and Gas industry. Augmentation methods such as CycleGans are also included, while experiments highlight the value of transformers outperforming convolutional neural networks both in pipes and vessels and more efficient detection of corrosion on pipes and inside vessels with high Intersection over Union (IoU).