This study aims to classify lithology in the Poseidon gas field using convolutional neural networks (CNN). The study uses well logs from 3 wells, including gamma ray, density, and neutron porosity logs. The CNN model is trained on lithology data from one well and tested on a second well. Several approaches are taken to improve the accuracy of lithology prediction, including simplifying the lithology classes, adding derived well log inputs, and removing thin layers from the data. The best prediction accuracy achieved is 57.7%. The CNN method is found to be less accurate for thin layers and more training data may be needed. Additional preprocessing and increasing the model complexity could further improve prediction.