Infrared Thermography (IRT) is now an important tool in the realm of Non- Destructive Testing (NDT) and monitoring in different industries. The automated analysis of images obtained through IRT, employing conventional Deep Learning approaches, is hindered by the scarcity of data. Unlike RGB images, where there is an abundance of training data available, IRT images are limited in number, expensive to acquire, and very specific to environmental conditions. This makes conventional data-hungry Convolutional Neural Networks (CNNs) unsuitable for IRT applications because of overfitting. This thesis attempts to bridge the gap by exploring the possibility of One-Shot Learning (OSL) approaches in computer vision tasks related to IRT images.