Seismic Image Retrieval and Classification with Novel Slice Shuffling Data Augmentation

This paper introduces a framework for automating the analysis of seismic images, which is essential for geological studies. Leveraging the intersection of Computer Vision and Geology, it addresses the scarcity of automated solutions in this domain. Developed within the Reflection Connection challenge, the framework facilitates query-based retrieval of seismic images and identifying structural features. Techniques such as fine-tuning pre-trained architectures and employing one-shot classification methods were explored. Additionally, a novel data augmentation method, Slice Shuffling Augmentation for Geological features that enhanced the model performance was developed.