This study presents an attempt to apply cross-domain transfer learning to preliminary seismic facies classification in a seismic dataset situated in Palawan, Philippines. The focus is to adapt a deep learning model trained on the Netherlands F3 seismic survey dataset to a different geological setting of the Palawan dataset. Seismic imaging process allows for the investigation of the substrata to be able to produce geological insights that are crucial in hydrocarbon explorations. This work involves implementing classification and semantic segmentation of seismic facies using an assembled RES-UNet-FCN deep neural network model. The methodology is comprised of initial classification using a feature extractor, appending this feature extractor to form a segmentation model, and the final application of the trained model to the Palawan dataset using transfer learning techniques. The challenges of applying and fine-tuning the model to the new dataset are examined, particularly in situations with limited data inputs. The results demonstrate the effectiveness of the model in classifying the seismic facies within the Netherlands F3 dataset, while also encountering challenges in adapting to the more complex geological features of the Palawan dataset. The findings suggest that while neural networks are adept at identifying patterns within images, in the case of seismic datasets, improvements could be made when applying them to datasets with significant geological variation and complexity. The paper contributes to the broader understanding of applying machine learning techniques in geological interpretations, particularly in enhancing the efficiency and accuracy of seismic data analysis in hydrocarbon exploration.