Object detection algorithms to identify skeletal components in carbonate cores

Identification of constituent grains in carbonate rocks requires specialist experience. A carbonate sedimentologist must be able to distinguish between skeletal grains that change through geological ages, preserved in differing alteration stages, and cut in random orientations across core sections. Recent studies have demonstrated the effectiveness of machine learning in classifying lithofacies from thin section, core, and seismic images, with faster analysis times and reduction of natural biases. In this study, we explore the application and limitations of convolutional neural network (CNN) based object detection frameworks to identify and quantify multiple types of carbonate grains within close-up core images of carbonate lithologies. We compiled nearly 400 images of high-resolution core images from three ODP and IODP expeditions. Over 9000 individual carbonate components of 11 different classes were manually labelled from this dataset. Using pre-trained weights, a transfer learning approach was applied to evaluate one-stage (YOLO v5) and two-stage (Faster R–CNN) detectors under different feature extractors (CSP-Darknet53 and ResNet50-FPN, respectively). Despite the current popularity of one-stage detectors, our results show Faster R–CNN with ResNet50-FPN backbone provides the most robust performance, achieving 0.73 mean average precision (mAP). Furthermore, we extend the approach by deploying the trained model to two ODP sites from Leg 194 that were not part of the training set (ODP Sites 1196 and 1199), providing a performance comparison with benchmark human interpretation.