# Comparing-CNN-and-ViT-in-Geological-Fault-Detection
Code for paper "A Comparative Study of Vision Transformer and Convolutional Neural Network Models in Fault Detection"
| filename | usage |
|---|---|
| models | including 6 models:unet, resnet50unet,transunet transattunet,swinunet, and swindeeplay |
| dataset | FaultsegDataset:FaultSeg3D ThebeDataset:Thebe RandSelectThebe:random select Thebe |
| configs | config file of swin transformer |
| dataefficient | predit3DThebe.py: the process of using 2d model to predict 3D cube |
| losses | loss file |
| utils | utility of the project, including handle images, models and metrics |
| model_evaluation | evaluate_thebe.py/evaluate_faultseg:get metrics results visual_Thebe/FaultSeg:visual predict fault lines predict_kerry3D:predict kerry3D fault predict3DThebe:predict Thebe and recover to 3D cube |
| warmup_schedule | warmup file |
| optins.py | config file |
| train_Faultseg.py | train Faultseg |
| train_Thebe.py | train Thebe |